VADD · Data generation, identifiable action anchors, and qualitative evidence
01
Change the requested action while preserving source content that the edit does not target.
THEORY invariance and action informationMETHOD source-conditioned action adaptersINFERENCE self-generated demonstrations
VADD learns action anchors from demonstrations and trains an editor to execute them under source conditioning. The data-generating process, identification result, method, and video evidence follow.
IMotivation
IIIdentifiable action anchors
IIIVADD editor
IVQualitative results
I · MOTIVATION
Change What Happens—Preserve Who and What
An action edit is valid only when the requested event occurs with the source-owned subjects and objects.
02
FM-1 · NO EDIT
The requested event never starts
The model is unresponsive: the source action continues, or the edit arrives only after the incompatible source event has already been replayed.
FM-2 · INVENTORY / INTERACTION
The object appears on demand
A manipulated object materializes, disappears, or bypasses the intended physical relation instead of remaining the same source-owned instance.
FM-3 · SUBJECT / ROLE REPLACEMENT
The interaction uses the wrong identity
Motion occurs, but the original actor or patient is replaced, re-rendered, or rebound. Plausible interaction cannot compensate for subject-preservation failure.
Non-compensatory gates
Action response, same-instance continuity, interaction topology, and subject preservation must all pass.
I · MOTIVATION · SELF-GENERATED DEMONSTRATIONS
Recover Action Capability from Demonstrations
Self-generated anchors reveal missing action mechanics; the editor should add only their source-faithful counterparts.
03
Conceptual support-space view. Shapes encode capability-set relations only; their area, distance, and anchor count do not represent probability mass.
Provisional design01 · BASE EVIDENCE
Generation competence ≠ editing competence
Orange anchors show that frozen T2V can generate an action mode that source-conditioned editing may still miss.
02 · TRANSFER RULE
The anchor is evidence—not the target
Its actor, objects, scene, and camera may differ. Only qualified event mechanics can transfer.
03 · LOCAL ADDITION
Recover the source-faithful counterpart
The recovered action must express the requested behavior while preserving source-owned identity, inventory, background, and unrelated motion.
I · MOTIVATION · DIRECT FAILURE EVIDENCE
Failure Mode 1 — The Requested Event Never Starts
Source-conditioned output can preserve source motion even when an action-only anchor makes the requested edit explicit.
04
Established evidence
Jellyfish drift → bell contraction
REQUESTED EVENT NEVER STARTS
Jellyfish drift → bell contraction
Real target contracts into a ring and the anchor demonstrates strong bell-shape change. Frozen keeps the source jellyfish rounded and drifting; contraction never begins.
Source · rounded drift
Real target · contracts to ring
T2V anchor · motion evidence
Frozen base · rounded drift persists
Ordinary phone use → playful selfie
REQUESTED EVENT NEVER STARTS
Playful selfie
Target and anchor lift the phone and add tongue/wink behavior; frozen continues ordinary phone use.
Source
Real target
T2V anchor
Frozen base
Direct evidence for FM-1
Bell contraction and the playful selfie exist in the target and action-anchor branches, but neither requested transition starts reliably in frozen source-conditioned editing.
I · MOTIVATION · FULL-SEQUENCE EVIDENCE
A Late Target-Like Endpoint Is Not an Edit
Frozen may reach part of the requested endpoint only after replaying most of the incompatible source sequence.
05
Established evidence
Dance in place → full scarf spin
DELAYED EDIT
The spin arrives after the dance
Frozen replays source-like in-place scarf motion for most of the clip, then begins a late turn. A late spin does not replace the earlier source event.
Source · dances in place
Real target · full spin
T2V anchor · spin evidence
Frozen base · late turn
One-arm face cover → eye rub and two-hand cover
DELAYED EDIT
The second hand arrives at the end
Frozen holds the source one-arm cover through most of the sequence, then adds the second hand late. Ordered eye-rubbing motion remains absent.
Source · one-arm cover
Real target · rub then cover
T2V anchor · motion evidence
Frozen base · late two-hand cover
Onset and order are non-compensatory
Sparse endpoint frames can look compliant while the full video remains source-driven. Later target similarity cannot erase an incompatible event that already occurred.
I · MOTIVATION · MANUAL FAILURE EVIDENCE
Failure Mode 1b — Source Motion Is Replayed Before the Edit
A delayed or partial response is still failure when the incompatible source event has already occurred.
06
Established evidence
Turn and place the bottle on the treadmill
SOURCE MOTION IS REPLAYED
Bottle placement
Target and anchor turn and place the bottle; frozen keeps holding it and facing forward.
Source
Real target
T2V anchor
Frozen base
Bow → stand upright and clap
SOURCE MOTION IS REPLAYED
Bow → clap
Frozen first inherits the complete bow, then only partially brings the hands together near the end.
Source
Real target
T2V anchor
Frozen base
Source-motion leakage
These frozen-editor examples retain the original source action. Check action execution together with subject and scene preservation.
I · MOTIVATION · PAPER AUDIT
Self-Generated Demonstrations vs. Edited Outputs
128 source–target pairs · frozen Bernini-R 1.3B
07
Ordered V-JEPA2 action margin and pairwise preference
Candidate
Mean
Metric preferred
Human preferred
Self-generated T2V
−0.01031
125 / 128
114 / 128
Source-conditioned edit
−0.14660
3 / 128
12 / 128
Tie
—
0 / 128
2 / 128
PAPER UPDATE · 01 / 06 · DATA GENERATING PROCESS
Data Generating Process: Content × Action
VADD: Video Action Editing via Demonstrations and Disentanglement
08
GENERATE
Roll out an ordered trajectory
Content is independent of the initial state, action regime , and innovations. The renderer combines content with the full realized trajectory .
EDIT
Recombine content and action
Preserve source content; transfer the demonstration’s trajectory, including its initial state and timing, within a fixed physical context.
SIMULATE
MuJoCo supplies controlled views
Replay physical motion under two block appearances. Physics evolves the objects; a fixed camera and lighting render each clip.
§2.1 + Appendix D · The strict controlled model fixes physical context; real consecutive-event clips approximate this coupling. Manuscript source ↗
PAPER UPDATE · 02 / 06 · IDENTIFIABILITY
When Does an Action Anchor Identify Action?
The positive pair shares the entire realized trajectory, not only the action label.
09
ACTION-PRESERVING VIEWS
POPULATION OBJECTIVE
Agreement removes appearance variation. Prediction of a characteristic kernel embedding retains differences between action-conditional video distributions.
Conditions for the theorem
Regularity: compact supports and continuous dynamics / rendering.
Appearance variation: independent, full-support content resampling, independent of the shared trajectory.
Temporal observability: distinct trajectories cannot produce the same ordered video under supported content choices.
Optimize over continuous encoders, finite-risk predictors, and a code dimension admitting continuous injective trajectory coordinates.
AT EVERY POPULATION GLOBAL OPTIMUM
is injective on a full-measure action subset; is the action-conditional kernel mean embedding. Individual semantic coordinates are not fixed.
§2.2 · Original manuscript equations. LaTeX source ↗ · Finite-data optimization and real-video patch views are outside this guarantee. Manuscript source ↗
PAPER UPDATE · 03 / 06 · TEMPORAL INFORMATION
Why the Ordered Video Matters
Three observations identify position, velocity, and acceleration; two can remain ambiguous.
10
NONLINEAR OBSERVATIONS
Temporal observability concerns the whole ordered clip. A single frame need not reveal the full action state.
Why ordinary mean prediction can fail
The multiplicative map with symmetric nonzero content is invertible, yet . Predicting this mean can erase all action differences.
Manuscript numerical check ·
Mixing / objective
Content
Shear · agreement + kernel
1.00
1.00
0.97
0.00
Multiplicative · same objective
1.00
1.00
1.00
0.00
When feature means suffice
In the additive-feature corollary, action differences must survive outside the appearance-variation subspace. Ordinary feature prediction then retains action.
§2.3 + Appendix E · values are three-seed means from the paper’s numerical experiment, separate from rendered-video examples. Manuscript source ↗
PAPER UPDATE · 04 / 06 · REPRESENTATION LEARNING
Learning Action Anchors from Demonstrations
A temporal model keeps ordered dynamics across spatial views of the same demonstration.
11
01
Observe dynamic features
Frozen features → temporal centering, signed changes, and local change magnitudes. Alternating patches supply two views.
02
Encode an action sequence
A bidirectional GRU predicts Gaussian posteriors. Their means are the action anchors; a recurrent, history-conditioned prior regularizes them.
§3.1 · Same-scene patch views and mean-target SmoothL1 are practical surrogates for the population criterion. The prior is a temporal inductive bias. Manuscript source ↗
PAPER UPDATE · 05 / 06 · EDITOR
Source-Conditioned Action Injection
Action specifies the motion; the source provides the appearance and spatial context in which it is executed.
12
DEMONSTRATION Normalized anchor
learned ordered action
+DEMONSTRATION Compressed V-JEPA2
complementary behavior features
→DETACHED CONDITION
phase-aligned feature concatenation
SOURCE-CONDITIONED LOW-RANK UPDATE
GENERATED STATE ·
Adapt to the current token
The first term couples the demonstrated action to the current generated state.
SOURCE CONTEXT ·
Bind the action to the source
The second term modulates the action using spatial context from independently encoded source frames.
SAMPLING GATE ·
Inject early, refine later
Adapters act on visual self-attention value / output projections during the first 8 of 40 sampling steps.
Scope of the theory
The identified object is the learned mean code. Normalization can discard offsets and scale; the augmented packet also includes V-JEPA2 features.
§3.2 + Appendix D · Frozen backbone and representation model; trainable residual adapters. Manuscript source ↗
PAPER UPDATE · 06 / 06 · TRAINING & INFERENCE
Train on Behavior; Preserve Source Content
Supervise the decoded clean prediction with ordered behavior and source-preservation signals.
13
ORDERED BEHAVIOR READER
+ Demonstration− Source− Reversed− Shuffled
The frozen V-JEPA2 reader differentiates through generated videos. Demonstration and control descriptors are detached.
Source preservation
Compare patch sets and appearance statistics. Remove the gradient component opposing behavior, then cap preservation RMS at ¼ of the behavior-gradient RMS without amplifying it.
Training
1,508 real source–target pairs; real-target and offline-generated demonstrations at 3:1. Action-code warm-up precedes behavior supervision.
Inference on OmniBench
Source + instruction → Qwen3-VL rewrites a T2V prompt.
Frozen Bernini generates the demonstration.
Extract once and run the conditional sampler.
No real target is read; inference needs neither behavior-reader gradients nor per-example optimization.
§3.3 + §4.1 / Appendix D · Warm-up followed by behavior supervision; inference uses a self-generated demonstration. Manuscript source ↗
EXPERIMENTS · QUANTITATIVE CONTEXT
OmniEdit-Bench Temporal: Paper Results
Accuracy-gated scores from the manuscript · 0–100, higher is better · category counts in parentheses
14
Method
Motion semantic (30)
Motion attribute (50)
Temporal composition (70)
Action (150)
Camera (50)
All (200)
VACE
6.80
7.54
11.48
9.23
8.00
8.92
ReCo
5.59
3.14
6.46
5.18
16.43
7.99
OmniVideo2
32.58
13.15
12.64
16.80
14.77
16.29
Kiwi-Edit
29.31
21.70
19.31
22.11
15.39
20.43
LiveEdit
18.80
13.92
10.79
13.44
7.32
11.91
Bernini · frozen
31.36
21.54
13.46
19.73
42.99
25.55
VADD · ours
34.93
24.96
17.64
23.54
46.16
29.19
Direct-FM · reference
38.23
14.56
15.85
19.90
25.82
21.38
ACTION
+3.81 over Bernini
23.54 vs. 19.73 across 150 non-camera cases.
ALL TEMPORAL
+3.64 over Bernini
29.19 vs. 25.55 across all 200 cases.
READING THE NEXT TWO CASES
Selected qualitative examples
High-score cases are a showcase; the table summarizes the complete evaluation.
§4, Omni Temporal table · Direct-FM is a paired-target reference and is not ranked. The following archive scores use Qwen3-VL-32B / 12 frames. Manuscript source ↗
QUALITATIVE · OMNIBENCH
Delete the flying kiss
Delete the action of the girl giving a flying kiss, while maintaining the continuity of the video naturally.
Clips start together at their original speed. Shorter clips hold their last frame; use the full video link for individual controls.
Appearance × trajectory
Preserve source colors while following the reference motion. Compare the approach, direction changes, and final placement at 1× speed.
View sampled frames and audit records
Time →Frame 0Frame 27Frame 53Frame 80
Source
Reference
Ours
Expected · audit only
Archived source record
Archived self-reconstruction
Uses the manuscript’s author-confirmed my/source pairing. The paper’s provenance note records the older source archive as incorrect. Expected frames are audit-only. Pairing record ↗
QUALITATIVE · SYNTHETIC
Push forward-left → Push back-right
Keep the source block appearance while following the demonstrated trajectory.
Clips start together at their original speed. Shorter clips hold their last frame; use the full video link for individual controls.
Appearance × trajectory
Preserve source colors while following the reference motion. Compare the approach, direction changes, and final placement at 1× speed.
View sampled frames and audit records
Time →Frame 0Frame 27Frame 53Frame 80
Source
Reference
Ours
Expected · audit only
Archived source record
Archived self-reconstruction
Uses the manuscript’s author-confirmed my/source pairing. The paper’s provenance note records the older source archive as incorrect. Expected frames are audit-only. Pairing record ↗
QUALITATIVE · REAL
Holding a camera → Looking through it
Edit the action so that the woman raises her silver camera with both hands, bringing it up to her eye to look through the viewfinder.
两模型共用的最终 caption 正文（无扩写）： The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.
两模型共用的最终 caption 正文（无扩写）： The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.
两模型共用的最终 caption 正文（无扩写）： The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.
两模型共用的最终 caption 正文（无扩写）： The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.
两模型共用的最终 caption 正文（无扩写）： The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.
两模型共用的最终 caption 正文（无扩写）： The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.
两模型共用的最终 caption 正文（无扩写）： The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.
Editing instruction：Edit the action so that the woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking.
两模型共用的最终 caption 正文（无扩写）： The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.
两模型共用的最终 caption 正文（无扩写）： The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.
两模型共用的最终 caption 正文（无扩写）： The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.
两模型共用的最终 caption 正文（无扩写）： The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.
两模型共用的最终 caption 正文（无扩写）： The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.
d908552fb691 · seed 2026081903 · caption SHA-256: 025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896
\ No newline at end of file
diff --git a/projects/video-action-editing/large-fair12/inputs.json b/projects/video-action-editing/large-fair12/inputs.json
new file mode 100644
index 0000000..027c230
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/inputs.json
@@ -0,0 +1,162 @@
+{
+ "schema_version": "large-fair12-fixed-caption-v1",
+ "models": [
+ "ByteDance/Bernini-R-Diffusers",
+ "Fudan-FUXI/OmniVideo2-A14B"
+ ],
+ "arms": [
+ "t2v",
+ "editing"
+ ],
+ "caption_policy": "Shared final body, target action verbatim followed by manually reviewed static identity/scene context. No automatic caption generation or rewriting. Native fixed task prefixes and feature extraction retained and recorded.",
+ "caption_context_source": "Existing MEV metadata reviewed to remove non-target actions; no target video passed to inference.",
+ "frames": 81,
+ "fps": 25,
+ "steps": 40,
+ "cases": [
+ {
+ "index": 0,
+ "case_id": "10af6c651dbc",
+ "name": "Jellyfish contraction",
+ "seed": 2026081910,
+ "original_instruction": "Edit the action so that the jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water.",
+ "target_action_caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water.",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/10af6c651dbc/source-exact81.mp4",
+ "source_sha256": "ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb"
+ },
+ {
+ "index": 1,
+ "case_id": "7fa338474c75",
+ "name": "Playful selfie",
+ "seed": 2026081900,
+ "original_instruction": "Edit the action so that the woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking.",
+ "target_action_caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking.",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7fa338474c75/source-exact81.mp4",
+ "source_sha256": "4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840"
+ },
+ {
+ "index": 2,
+ "case_id": "81533c9e56ec",
+ "name": "Scarf spin",
+ "seed": 2026081915,
+ "original_instruction": "Edit the action so that the woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf.",
+ "target_action_caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf.",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/81533c9e56ec/source-exact81.mp4",
+ "source_sha256": "6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19"
+ },
+ {
+ "index": 3,
+ "case_id": "d977ce147d80",
+ "name": "Rub eyes then cover face",
+ "seed": 2026081905,
+ "original_instruction": "Edit the action so that the man lying on the couch rubs his tired eyes and covers his face with both hands.",
+ "target_action_caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands.",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d977ce147d80/source-exact81.mp4",
+ "source_sha256": "e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d"
+ },
+ {
+ "index": 4,
+ "case_id": "840b214afead",
+ "name": "Place bottle on treadmill",
+ "seed": 2026081913,
+ "original_instruction": "Edit the action so that the woman turns her head to the left and places her water bottle onto the treadmill.",
+ "target_action_caption": "The woman turns her head to the left and places her water bottle onto the treadmill.",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/840b214afead/source-exact81.mp4",
+ "source_sha256": "a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646"
+ },
+ {
+ "index": 5,
+ "case_id": "d908552fb691",
+ "name": "Bow to clap",
+ "seed": 2026081903,
+ "original_instruction": "Edit the action so that the woman in a taekwondo uniform stands in profile and claps her hands together in front of her.",
+ "target_action_caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her.",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d908552fb691/source-exact81.mp4",
+ "source_sha256": "56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb"
+ },
+ {
+ "index": 6,
+ "case_id": "fb7725c351ba",
+ "name": "提杠铃 · Barbell lift",
+ "seed": 2026081901,
+ "original_instruction": "Edit the action so that the woman in a home gym lifts a barbell from the floor to her waist.",
+ "target_action_caption": "The woman in a home gym lifts a barbell from the floor to her waist.",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/fb7725c351ba/source-exact81.mp4",
+ "source_sha256": "479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a"
+ },
+ {
+ "index": 7,
+ "case_id": "3fadbe0c2961",
+ "name": "狗转头 · Dog head turn",
+ "seed": 2026081902,
+ "original_instruction": "Edit the action so that the dog turns its head further to the right while sitting in the grass.",
+ "target_action_caption": "The dog turns its head further to the right while sitting in the grass.",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/3fadbe0c2961/source-exact81.mp4",
+ "source_sha256": "99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364"
+ },
+ {
+ "index": 8,
+ "case_id": "7077cfb160cf",
+ "name": "长椅上写字 · Write in notebook",
+ "seed": 2026081904,
+ "original_instruction": "Edit the action so that the young woman sitting on a park bench thoughtfully writes in her notebook with a pen.",
+ "target_action_caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen.",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7077cfb160cf/source-exact81.mp4",
+ "source_sha256": "cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9"
+ },
+ {
+ "index": 9,
+ "case_id": "7e488525099e",
+ "name": "转身倚车 · Turn and lean on car",
+ "seed": 2026081906,
+ "original_instruction": "Edit the action so that the woman in a black leather jacket and red pants turns and leans back against the hood of the black car.",
+ "target_action_caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car.",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7e488525099e/source-exact81.mp4",
+ "source_sha256": "1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4"
+ },
+ {
+ "index": 10,
+ "case_id": "d5221a1b69d5",
+ "name": "闭眼尝汤 · Taste from ladle",
+ "seed": 2026081907,
+ "original_instruction": "Edit the action so that the woman in an apron closes her eyes as she tastes from a metal ladle.",
+ "target_action_caption": "The woman in an apron closes her eyes as she tastes from a metal ladle.",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d5221a1b69d5/source-exact81.mp4",
+ "source_sha256": "fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14"
+ },
+ {
+ "index": 11,
+ "case_id": "40712e1341dc",
+ "name": "放下手机后翻身睡觉 · Put phone away and sleep",
+ "seed": 2026081911,
+ "original_instruction": "Edit the action so that the man in bed puts his smartphone on the nightstand and turns over to go to sleep.",
+ "target_action_caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep.",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/40712e1341dc/source-exact81.mp4",
+ "source_sha256": "654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69"
+ }
+ ]
+}
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_editing.jpg
new file mode 100644
index 0000000..216f07c
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_editing.json b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_editing.json
new file mode 100644
index 0000000..fc251ef
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "10af6c651dbc",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 104,
+ "prompt_truncated": false,
+ "seed": 2026081910,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/10af6c651dbc/source-exact81.mp4",
+ "source_sha256": "ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "sigma_min",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "82c8810cd7546c2d6ea80743c3c90d4a926e37f5099790e2b6acf5fc19c77510",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:36:06.353806+00:00",
+ "elapsed_seconds": 1623.09,
+ "job": "214745",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_editing.mp4
new file mode 100644
index 0000000..f2a0d05
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_t2v.jpg
new file mode 100644
index 0000000..32acb1a
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_t2v.json
new file mode 100644
index 0000000..c645768
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "10af6c651dbc",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 65,
+ "prompt_truncated": false,
+ "seed": 2026081910,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "99c080cb8ebc719e47354dbc79ba641c7ffcc30c355bbafd05335903314ba8d8",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:04:25.056432+00:00",
+ "elapsed_seconds": 579.95,
+ "job": "214700",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_t2v.mp4
new file mode 100644
index 0000000..eb5b331
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/10af6c651dbc/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_editing.jpg
new file mode 100644
index 0000000..a899257
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_editing.json b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_editing.json
new file mode 100644
index 0000000..b1c5777
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "10af6c651dbc",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "actual_t5_text": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "10af6c651dbc",
+ "arm": "editing",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/10af6c651dbc/source-exact81.mp4",
+ "source_sha256": "ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 74,
+ 2048
+ ],
+ "feature_sha256": "31d5a8e438020fe75f1d26e7ad419c5e28ba4983c9e49203dd1a52efd9017d75"
+ },
+ "t5_text_inputs": [
+ "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081910,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/10af6c651dbc/source-exact81.mp4",
+ "source_sha256": "ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "d7a3b42e059c8ac6133f96b2fc4664509d504c9f821f45cc24d9d99acb151d44",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:56:21.408682+00:00",
+ "elapsed_seconds": 921.01,
+ "job": "214733",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_editing.mp4
new file mode 100644
index 0000000..99cfa78
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_t2v.jpg
new file mode 100644
index 0000000..3f7572e
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_t2v.json b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_t2v.json
new file mode 100644
index 0000000..08ef732
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "10af6c651dbc",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "actual_t5_text": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "10af6c651dbc",
+ "arm": "t2v",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 47,
+ 2048
+ ],
+ "feature_sha256": "b1d9193dbd160064ceb99196242d2707a2f74e510993302982dac03e6b713c01"
+ },
+ "t5_text_inputs": [
+ "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081910,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "1d1edd0eef41ff1fcde1a1bd662fdb47694a3b9ed60e458e83e842af441f9eb6",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:18:02.505227+00:00",
+ "elapsed_seconds": 608.46,
+ "job": "214721",
+ "hostname": "auh7-1b-gpu-312",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_t2v.mp4
new file mode 100644
index 0000000..fa25e1d
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/10af6c651dbc/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_editing.jpg
new file mode 100644
index 0000000..7b22911
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_editing.json b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_editing.json
new file mode 100644
index 0000000..3c2fd57
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "3fadbe0c2961",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 102,
+ "prompt_truncated": false,
+ "seed": 2026081902,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/3fadbe0c2961/source-exact81.mp4",
+ "source_sha256": "99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "sigma_max",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "59ddf05fd48e62dffbc27aac5a721cb7198c40d526c06479056695547d9cf48f",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:49:23.038345+00:00",
+ "elapsed_seconds": 1621.77,
+ "job": "214751",
+ "hostname": "auh7-1b-gpu-307",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_editing.mp4
new file mode 100644
index 0000000..f25eb23
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_t2v.jpg
new file mode 100644
index 0000000..46f6315
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_t2v.json
new file mode 100644
index 0000000..2938bca
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "3fadbe0c2961",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 63,
+ "prompt_truncated": false,
+ "seed": 2026081902,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "dc894ddffe1960fa0a845b2bbf03c37162d8b496e887ef535bed5cbf2ace9616",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:45:29.419504+00:00",
+ "elapsed_seconds": 706.21,
+ "job": "214706",
+ "hostname": "auh7-1b-gpu-269",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_t2v.mp4
new file mode 100644
index 0000000..aad94a1
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_editing.jpg
new file mode 100644
index 0000000..385ab6c
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_editing.json b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_editing.json
new file mode 100644
index 0000000..585ccaa
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "3fadbe0c2961",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "actual_t5_text": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "3fadbe0c2961",
+ "arm": "editing",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/3fadbe0c2961/source-exact81.mp4",
+ "source_sha256": "99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 78,
+ 2048
+ ],
+ "feature_sha256": "5607250394671775f7a81e3929b7e9ff08652121e7def64d0a370240e1f6e2ce"
+ },
+ "t5_text_inputs": [
+ "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081902,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/3fadbe0c2961/source-exact81.mp4",
+ "source_sha256": "99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "87aa2a665e5c451327950cf7eaa217e0f10ba00a2dac3f85af22489eb046501b",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:07:32.501671+00:00",
+ "elapsed_seconds": 637.79,
+ "job": "214739",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_editing.mp4
new file mode 100644
index 0000000..f524ffe
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_t2v.jpg
new file mode 100644
index 0000000..bbf38da
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_t2v.json b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_t2v.json
new file mode 100644
index 0000000..d884a57
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "3fadbe0c2961",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "actual_t5_text": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "3fadbe0c2961",
+ "arm": "t2v",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 51,
+ 2048
+ ],
+ "feature_sha256": "13a1edc8e7db94239ea6c6049d5691001949d9c95c3900fb5bbfd9298401f18a"
+ },
+ "t5_text_inputs": [
+ "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081902,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "e92d7a1bcddf3582e33dbd1b93244e9a12de37a18ea27ffff2088f026e3584ec",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:35:53.507686+00:00",
+ "elapsed_seconds": 637.05,
+ "job": "214727",
+ "hostname": "auh7-1b-gpu-307",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_t2v.mp4
new file mode 100644
index 0000000..65c4b34
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/3fadbe0c2961/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_editing.jpg
new file mode 100644
index 0000000..88b19c1
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_editing.json b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_editing.json
new file mode 100644
index 0000000..69ce857
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "40712e1341dc",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 113,
+ "prompt_truncated": false,
+ "seed": 2026081911,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/40712e1341dc/source-exact81.mp4",
+ "source_sha256": "654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_max",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "d6b8387a8f977f88fec1fa6f16ef0ad0ae9c22554a74bb661014ad6260532037",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:30:33.960692+00:00",
+ "elapsed_seconds": 1605.68,
+ "job": "214755",
+ "hostname": "auh7-1b-gpu-283",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_editing.mp4
new file mode 100644
index 0000000..97b0c82
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_t2v.jpg
new file mode 100644
index 0000000..d0408e2
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_t2v.json
new file mode 100644
index 0000000..9fb1449
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "40712e1341dc",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 74,
+ "prompt_truncated": false,
+ "seed": 2026081911,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "sigma_max",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "ac158010a45d2c746774740d9a513440d29ac8d82604895d75b41c73e1461eef",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:45:13.322144+00:00",
+ "elapsed_seconds": 482.84,
+ "job": "214710",
+ "hostname": "auh7-1b-gpu-283",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_t2v.mp4
new file mode 100644
index 0000000..268a24c
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/40712e1341dc/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_editing.jpg
new file mode 100644
index 0000000..2fdb075
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_editing.json b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_editing.json
new file mode 100644
index 0000000..6de9c88
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "40712e1341dc",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "actual_t5_text": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "40712e1341dc",
+ "arm": "editing",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/40712e1341dc/source-exact81.mp4",
+ "source_sha256": "654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 88,
+ 2048
+ ],
+ "feature_sha256": "17f26bdae837dadcbf1c11b139c2c9ba8cbaa805f2f35c4fff16b1a5a0a45f67"
+ },
+ "t5_text_inputs": [
+ "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081911,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/40712e1341dc/source-exact81.mp4",
+ "source_sha256": "654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "271248625a365fabc0664f1e5f3cfadd94ad2179e17a2efa261b9d5926d3765b",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:20:56.467352+00:00",
+ "elapsed_seconds": 646.24,
+ "job": "214743",
+ "hostname": "auh7-1b-gpu-307",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_editing.mp4
new file mode 100644
index 0000000..5f5cff0
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_t2v.jpg
new file mode 100644
index 0000000..86764ce
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_t2v.json b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_t2v.json
new file mode 100644
index 0000000..9abf792
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "40712e1341dc",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "actual_t5_text": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "40712e1341dc",
+ "arm": "t2v",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 61,
+ 2048
+ ],
+ "feature_sha256": "f953db0ef97435bed85eb46eb0a25136e1775f7d4342e1a435a9ed189580145c"
+ },
+ "t5_text_inputs": [
+ "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081911,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "9dafe547fd5b649dccbf097319e9f41141518721416a7a75534548dbf4f138e3",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:26:43.210982+00:00",
+ "elapsed_seconds": 958.87,
+ "job": "214756",
+ "hostname": "auh7-1b-gpu-210",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_t2v.mp4
new file mode 100644
index 0000000..6e2b258
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/40712e1341dc/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_editing.jpg
new file mode 100644
index 0000000..05bb617
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_editing.json b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_editing.json
new file mode 100644
index 0000000..eec2b61
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "7077cfb160cf",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 107,
+ "prompt_truncated": false,
+ "seed": 2026081904,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7077cfb160cf/source-exact81.mp4",
+ "source_sha256": "cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "sigma_min",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "2b0371a6e0660520daf895823f2f74b556503b075578fe2dca02f5c398bd23e7",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:05:32.304246+00:00",
+ "elapsed_seconds": 2298.97,
+ "job": "214752",
+ "hostname": "auh7-1b-gpu-210",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_editing.mp4
new file mode 100644
index 0000000..6d43d11
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_t2v.jpg
new file mode 100644
index 0000000..5e021da
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_t2v.json
new file mode 100644
index 0000000..cf5fe81
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "7077cfb160cf",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 68,
+ "prompt_truncated": false,
+ "seed": 2026081904,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "shift_terminal",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "eaf121f7e20fa7dc7aa6d4d127b00180cb92fee67e1fdea5ff2fbc7b9be7d997",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:19:58.997189+00:00",
+ "elapsed_seconds": 705.61,
+ "job": "214757",
+ "hostname": "auh7-1b-gpu-269",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_t2v.mp4
new file mode 100644
index 0000000..9e698ff
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7077cfb160cf/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_editing.jpg
new file mode 100644
index 0000000..7c08f9b
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_editing.json b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_editing.json
new file mode 100644
index 0000000..62093d0
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "7077cfb160cf",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "actual_t5_text": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7077cfb160cf",
+ "arm": "editing",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7077cfb160cf/source-exact81.mp4",
+ "source_sha256": "cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 80,
+ 2048
+ ],
+ "feature_sha256": "70de5d8bbce06459f31e50438e02db4aae638a7c14eec75b64875ee09eb289cb"
+ },
+ "t5_text_inputs": [
+ "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081904,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7077cfb160cf/source-exact81.mp4",
+ "source_sha256": "cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "3b158d997abb43e583b01b618a280bb0938570e81b73116ce351f2cb255e96af",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:08:05.180195+00:00",
+ "elapsed_seconds": 649.79,
+ "job": "214740",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_editing.mp4
new file mode 100644
index 0000000..287a44e
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_t2v.jpg
new file mode 100644
index 0000000..55d7703
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_t2v.json b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_t2v.json
new file mode 100644
index 0000000..580be72
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "7077cfb160cf",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "actual_t5_text": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7077cfb160cf",
+ "arm": "t2v",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 53,
+ 2048
+ ],
+ "feature_sha256": "863e64e016115388d775caa7fd1a8be61270b967361152418dfda43d47ba5cc6"
+ },
+ "t5_text_inputs": [
+ "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081904,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "cfce2698573f7b20a42686a08059ff6d39d7053cafcf4731fddab4d174c804ae",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:35:38.009473+00:00",
+ "elapsed_seconds": 599.71,
+ "job": "214728",
+ "hostname": "auh7-1b-gpu-283",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_t2v.mp4
new file mode 100644
index 0000000..d7c25db
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7077cfb160cf/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_editing.jpg
new file mode 100644
index 0000000..2064391
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_editing.json b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_editing.json
new file mode 100644
index 0000000..3d4e1b6
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "7e488525099e",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 97,
+ "prompt_truncated": false,
+ "seed": 2026081906,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7e488525099e/source-exact81.mp4",
+ "source_sha256": "1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "9829080372e3351bb3375d787615f72cc4a10be693f382edd9d36fae37079d0e",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:38:14.312525+00:00",
+ "elapsed_seconds": 2268.32,
+ "job": "214753",
+ "hostname": "auh7-1b-gpu-221",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_editing.mp4
new file mode 100644
index 0000000..b551660
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_t2v.jpg
new file mode 100644
index 0000000..6c5be54
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_t2v.json
new file mode 100644
index 0000000..47ebb9b
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "7e488525099e",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 58,
+ "prompt_truncated": false,
+ "seed": 2026081906,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "sigma_min",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "4d3b693a341087b216eba42bc9933ed61efbead441536c78638dfd4e5cca9145",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:17:05.780853+00:00",
+ "elapsed_seconds": 519.14,
+ "job": "214758",
+ "hostname": "auh7-1b-gpu-312",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_t2v.mp4
new file mode 100644
index 0000000..aeffef0
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7e488525099e/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_editing.jpg
new file mode 100644
index 0000000..e97f08d
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/omni_editing.json b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_editing.json
new file mode 100644
index 0000000..76b5c5a
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "7e488525099e",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "actual_t5_text": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7e488525099e",
+ "arm": "editing",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7e488525099e/source-exact81.mp4",
+ "source_sha256": "1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 71,
+ 2048
+ ],
+ "feature_sha256": "a34cbe81e1903e91618292d5ab323f49478f88c4638de853db8d3bb4475d4242"
+ },
+ "t5_text_inputs": [
+ "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081906,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7e488525099e/source-exact81.mp4",
+ "source_sha256": "1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "19321c8236ea55c3ce3006817c0c8316f9313bc193729e4ba6dd1651f445d5fc",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:15:57.054117+00:00",
+ "elapsed_seconds": 886.52,
+ "job": "214741",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_editing.mp4
new file mode 100644
index 0000000..00c0ea0
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_t2v.jpg
new file mode 100644
index 0000000..5c0c9c4
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/omni_t2v.json b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_t2v.json
new file mode 100644
index 0000000..11c925c
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "7e488525099e",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "actual_t5_text": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7e488525099e",
+ "arm": "t2v",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 44,
+ 2048
+ ],
+ "feature_sha256": "9f6d9a9afe61984fcb7355e65810471e2b4b623b73134d18a47daa0936b86b83"
+ },
+ "t5_text_inputs": [
+ "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081906,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "5d79f8d88fc14bccd0336603594e03abcc72413534b835b37698aed74b1bc7cf",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:36:39.023949+00:00",
+ "elapsed_seconds": 615.52,
+ "job": "214729",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7e488525099e/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_t2v.mp4
new file mode 100644
index 0000000..73a9231
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7e488525099e/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_editing.jpg
new file mode 100644
index 0000000..aedc14f
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_editing.json b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_editing.json
new file mode 100644
index 0000000..390d580
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "7fa338474c75",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 100,
+ "prompt_truncated": false,
+ "seed": 2026081900,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7fa338474c75/source-exact81.mp4",
+ "source_sha256": "4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "sigma_max",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "d80fbfb74a6ab3aed91b258c759a2eb21e37b66c82513742590b0e512e0f49f4",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:36:39.558813+00:00",
+ "elapsed_seconds": 1623.02,
+ "job": "214746",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_editing.mp4
new file mode 100644
index 0000000..679313d
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_t2v.jpg
new file mode 100644
index 0000000..d58e97f
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_t2v.json
new file mode 100644
index 0000000..b48c761
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "7fa338474c75",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 61,
+ "prompt_truncated": false,
+ "seed": 2026081900,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_max",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "9f68a732503e5d11d4b94ee86ea61175d555092ffc539b69615faf0e050a083d",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:03:56.015342+00:00",
+ "elapsed_seconds": 558.77,
+ "job": "214701",
+ "hostname": "auh7-1b-gpu-283",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_t2v.mp4
new file mode 100644
index 0000000..6f9ab7e
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7fa338474c75/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_editing.jpg
new file mode 100644
index 0000000..025caa3
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_editing.json b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_editing.json
new file mode 100644
index 0000000..499483e
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "7fa338474c75",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "actual_t5_text": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7fa338474c75",
+ "arm": "editing",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7fa338474c75/source-exact81.mp4",
+ "source_sha256": "4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 73,
+ 2048
+ ],
+ "feature_sha256": "dc6a8fcf6aace395826a95d45c233eabfd1bbacf9f833e8de912e13129a059d8"
+ },
+ "t5_text_inputs": [
+ "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081900,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7fa338474c75/source-exact81.mp4",
+ "source_sha256": "4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "e4267370c4b549bba46b2c14020f9fff2a725e1b4c90379041ab73bbeb59e8be",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:51:53.409740+00:00",
+ "elapsed_seconds": 636.88,
+ "job": "214734",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_editing.mp4
new file mode 100644
index 0000000..edab628
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_t2v.jpg
new file mode 100644
index 0000000..93528c7
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_t2v.json b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_t2v.json
new file mode 100644
index 0000000..08480a1
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "7fa338474c75",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "actual_t5_text": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7fa338474c75",
+ "arm": "t2v",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 46,
+ 2048
+ ],
+ "feature_sha256": "d54fddcf16850d14dc9dfb71b863ba619646a35c90cd4dd2d62419a2adda477a"
+ },
+ "t5_text_inputs": [
+ "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081900,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "0c6b53c4bae5fa1c7275fb71f5197c6312876d3813ae35b9c4db361712d3c65f",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:20:46.292401+00:00",
+ "elapsed_seconds": 595.85,
+ "job": "214722",
+ "hostname": "auh7-1b-gpu-283",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_t2v.mp4
new file mode 100644
index 0000000..9a584bb
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/7fa338474c75/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_editing.jpg
new file mode 100644
index 0000000..59614bf
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_editing.json b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_editing.json
new file mode 100644
index 0000000..ec1434d
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "81533c9e56ec",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 104,
+ "prompt_truncated": false,
+ "seed": 2026081915,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/81533c9e56ec/source-exact81.mp4",
+ "source_sha256": "6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "6568610950f2753ca3d4c6812018131d53241cb7921236502d6b4d4946380dd9",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:47:28.810955+00:00",
+ "elapsed_seconds": 2245.25,
+ "job": "214747",
+ "hostname": "auh7-1b-gpu-206",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_editing.mp4
new file mode 100644
index 0000000..d004ac6
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_t2v.jpg
new file mode 100644
index 0000000..c713c0d
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_t2v.json
new file mode 100644
index 0000000..659e9c7
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "81533c9e56ec",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 65,
+ "prompt_truncated": false,
+ "seed": 2026081915,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "sigma_max",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "d7a3869e2f1e3f30b90495f183234d1dcd56b5566c27678ac94669f243f45be8",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:03:06.079836+00:00",
+ "elapsed_seconds": 499.76,
+ "job": "214702",
+ "hostname": "auh7-1b-gpu-312",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_t2v.mp4
new file mode 100644
index 0000000..f7dc9d2
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/81533c9e56ec/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_editing.jpg
new file mode 100644
index 0000000..619992c
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_editing.json b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_editing.json
new file mode 100644
index 0000000..d62cae6
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "81533c9e56ec",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "actual_t5_text": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "81533c9e56ec",
+ "arm": "editing",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/81533c9e56ec/source-exact81.mp4",
+ "source_sha256": "6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 76,
+ 2048
+ ],
+ "feature_sha256": "09cbf7460835b917d1811c9612fb73eb0bd7c7ad17b988a8093c86702c180100"
+ },
+ "t5_text_inputs": [
+ "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081915,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/81533c9e56ec/source-exact81.mp4",
+ "source_sha256": "6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "d156e96a7fb7d91ea4a5e0159841362af432c9fed0b20882a8d508796398f671",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:52:22.199956+00:00",
+ "elapsed_seconds": 650.56,
+ "job": "214735",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_editing.mp4
new file mode 100644
index 0000000..8b2d9f1
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_t2v.jpg
new file mode 100644
index 0000000..466eaff
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_t2v.json b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_t2v.json
new file mode 100644
index 0000000..3bde35f
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "81533c9e56ec",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "actual_t5_text": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "81533c9e56ec",
+ "arm": "t2v",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 49,
+ 2048
+ ],
+ "feature_sha256": "55500ce71d10d808501849ae5c122f61dbff99b38bd16dd415fb146d599a62c3"
+ },
+ "t5_text_inputs": [
+ "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081915,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "a78f02ecb3aaa875cec8a133397227a526d4ea0e8dadf0a79bb3d7f8478f49e2",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:21:31.826216+00:00",
+ "elapsed_seconds": 612.01,
+ "job": "214723",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_t2v.mp4
new file mode 100644
index 0000000..cc30267
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/81533c9e56ec/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_editing.jpg
new file mode 100644
index 0000000..731d6e7
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/bernini_editing.json b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_editing.json
new file mode 100644
index 0000000..8a423b1
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "840b214afead",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 110,
+ "prompt_truncated": false,
+ "seed": 2026081913,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/840b214afead/source-exact81.mp4",
+ "source_sha256": "a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "sigma_max",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "aa4f8096577b9224c04eadfefb1581272b7a1355a9a730939813296ef15d39a2",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:59:20.369231+00:00",
+ "elapsed_seconds": 2401.86,
+ "job": "214749",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_editing.mp4
new file mode 100644
index 0000000..bc48556
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_t2v.jpg
new file mode 100644
index 0000000..cbad62f
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_t2v.json
new file mode 100644
index 0000000..d738a93
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "840b214afead",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 71,
+ "prompt_truncated": false,
+ "seed": 2026081913,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "sigma_min",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "88c0e40f239232c6cbb95be01496886f9100f26023b357bc13dd26715e4155cb",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:49:44.701325+00:00",
+ "elapsed_seconds": 545.02,
+ "job": "214704",
+ "hostname": "auh7-1b-gpu-307",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_t2v.mp4
new file mode 100644
index 0000000..c15dde8
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/840b214afead/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/840b214afead/omni_editing.jpg
new file mode 100644
index 0000000..87cbb9f
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/840b214afead/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/omni_editing.json b/projects/video-action-editing/large-fair12/results/840b214afead/omni_editing.json
new file mode 100644
index 0000000..be7e659
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/840b214afead/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "840b214afead",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "actual_t5_text": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "840b214afead",
+ "arm": "editing",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/840b214afead/source-exact81.mp4",
+ "source_sha256": "a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 81,
+ 2048
+ ],
+ "feature_sha256": "8d2079d5bd9b6692dfeb7a739b208a4eaf19cf1aca8b431b9a99ec76318349d8"
+ },
+ "t5_text_inputs": [
+ "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081913,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/840b214afead/source-exact81.mp4",
+ "source_sha256": "a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "8e7370fc84142be3daa707d6ffc7ff82556f7167a6fe14ed26ce5112d574f625",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:08:46.895702+00:00",
+ "elapsed_seconds": 974.76,
+ "job": "214737",
+ "hostname": "auh7-1b-gpu-206",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/840b214afead/omni_editing.mp4
new file mode 100644
index 0000000..fe04236
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/840b214afead/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/840b214afead/omni_t2v.jpg
new file mode 100644
index 0000000..e95202a
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/840b214afead/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/omni_t2v.json b/projects/video-action-editing/large-fair12/results/840b214afead/omni_t2v.json
new file mode 100644
index 0000000..75dbc9b
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/840b214afead/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "840b214afead",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "actual_t5_text": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "840b214afead",
+ "arm": "t2v",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 54,
+ 2048
+ ],
+ "feature_sha256": "dd48866a02a17042eef8b6cbde060576c10deaf6137ee965388861edfd103d7a"
+ },
+ "t5_text_inputs": [
+ "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081913,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "86baea5251e3c110b691d72b25cfb9405e3039affbbaf8e921c7e9169101b1c2",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:36:13.591878+00:00",
+ "elapsed_seconds": 1029.69,
+ "job": "214725",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/840b214afead/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/840b214afead/omni_t2v.mp4
new file mode 100644
index 0000000..17b4c85
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/840b214afead/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_editing.jpg
new file mode 100644
index 0000000..d0a6bae
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_editing.json b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_editing.json
new file mode 100644
index 0000000..9a2077e
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "d5221a1b69d5",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 92,
+ "prompt_truncated": false,
+ "seed": 2026081907,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d5221a1b69d5/source-exact81.mp4",
+ "source_sha256": "fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "shift_terminal",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "9cdbd07566d899630c7d460876abe733604cb04d284adf01842a9716a5db4ec3",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:38:07.139056+00:00",
+ "elapsed_seconds": 2248.64,
+ "job": "214754",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_editing.mp4
new file mode 100644
index 0000000..dfb4744
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_t2v.jpg
new file mode 100644
index 0000000..0975b64
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_t2v.json
new file mode 100644
index 0000000..911214c
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "d5221a1b69d5",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 53,
+ "prompt_truncated": false,
+ "seed": 2026081907,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "28120017a5e122b3f72956b9c2b16b20aae762baa4e88b9bab605ed03fff7c40",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:42:17.028942+00:00",
+ "elapsed_seconds": 483.92,
+ "job": "214709",
+ "hostname": "auh7-1b-gpu-312",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_t2v.mp4
new file mode 100644
index 0000000..dc54b5a
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_editing.jpg
new file mode 100644
index 0000000..e293a8a
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_editing.json b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_editing.json
new file mode 100644
index 0000000..d54db7e
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "d5221a1b69d5",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "actual_t5_text": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d5221a1b69d5",
+ "arm": "editing",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d5221a1b69d5/source-exact81.mp4",
+ "source_sha256": "fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 67,
+ 2048
+ ],
+ "feature_sha256": "c34f3a4e8b0058cf7a7f941082a5266c00ee63aea17fccbbdec08c7d9905eddb"
+ },
+ "t5_text_inputs": [
+ "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081907,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d5221a1b69d5/source-exact81.mp4",
+ "source_sha256": "fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "a1e43492995e5fefe7ace547ce6f62cb6fe363ae29ad365f7144f36f7bf4e900",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:23:04.739246+00:00",
+ "elapsed_seconds": 963.54,
+ "job": "214742",
+ "hostname": "auh7-1b-gpu-210",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_editing.mp4
new file mode 100644
index 0000000..50efe36
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_t2v.jpg
new file mode 100644
index 0000000..be49b6d
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_t2v.json b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_t2v.json
new file mode 100644
index 0000000..da9e1a5
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "d5221a1b69d5",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "actual_t5_text": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d5221a1b69d5",
+ "arm": "t2v",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 40,
+ 2048
+ ],
+ "feature_sha256": "32fb5857a4b6f85b7d8f9028089619067c830e83ddb787da525a2475cfdcbf7c"
+ },
+ "t5_text_inputs": [
+ "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081907,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "9ee4d254b5c2d579d3a9adcdc9d1057d7a8ef0afe1860f8a785db66d2cedd8ab",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:36:30.235433+00:00",
+ "elapsed_seconds": 603.1,
+ "job": "214730",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_t2v.mp4
new file mode 100644
index 0000000..f6cbcf2
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d5221a1b69d5/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_editing.jpg
new file mode 100644
index 0000000..d10e71d
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_editing.json b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_editing.json
new file mode 100644
index 0000000..505234b
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "d908552fb691",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 99,
+ "prompt_truncated": false,
+ "seed": 2026081903,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d908552fb691/source-exact81.mp4",
+ "source_sha256": "56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "shift_terminal",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "f164b04e01c335a49aca4f94494b87bd5ff2253a61f7d158fa2ddd00ade9967b",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:59:03.774298+00:00",
+ "elapsed_seconds": 2250.01,
+ "job": "214750",
+ "hostname": "auh7-1b-gpu-221",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_editing.mp4
new file mode 100644
index 0000000..489a8f3
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_t2v.jpg
new file mode 100644
index 0000000..4315027
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_t2v.json
new file mode 100644
index 0000000..4e97f48
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "d908552fb691",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 60,
+ "prompt_truncated": false,
+ "seed": 2026081903,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "shift_terminal",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "4316ede62030399da548d0b2b136333c622705fe1395b948ab7d60eca68430ce",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:45:46.786299+00:00",
+ "elapsed_seconds": 730.83,
+ "job": "214705",
+ "hostname": "auh7-1b-gpu-206",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_t2v.mp4
new file mode 100644
index 0000000..5886c73
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d908552fb691/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_editing.jpg
new file mode 100644
index 0000000..7a3cfb4
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/omni_editing.json b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_editing.json
new file mode 100644
index 0000000..1d5328c
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "d908552fb691",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "actual_t5_text": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d908552fb691",
+ "arm": "editing",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d908552fb691/source-exact81.mp4",
+ "source_sha256": "56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 78,
+ 2048
+ ],
+ "feature_sha256": "114740c8405aa14069531f4ff749cc0da9706f5fe07cf087a02f9e11bbdbc19c"
+ },
+ "t5_text_inputs": [
+ "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081903,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d908552fb691/source-exact81.mp4",
+ "source_sha256": "56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "aafdce633f1ac4a24120c74a0a75ddb3a4b8a4198e0734ce3643bf8124d8ceb2",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:05:24.214802+00:00",
+ "elapsed_seconds": 651.09,
+ "job": "214738",
+ "hostname": "auh7-1b-gpu-307",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_editing.mp4
new file mode 100644
index 0000000..6f8a96f
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_t2v.jpg
new file mode 100644
index 0000000..e28b181
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/omni_t2v.json b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_t2v.json
new file mode 100644
index 0000000..82cde1d
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "d908552fb691",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "actual_t5_text": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d908552fb691",
+ "arm": "t2v",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 51,
+ 2048
+ ],
+ "feature_sha256": "104ba746d40a2909261000fd07d3cefaf8f075beea1e2f6d4224a62c95937b01"
+ },
+ "t5_text_inputs": [
+ "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081903,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "7458de291881bf7e8eb7909a7dbad8fae1c01726304a6494eb637e4ba845c60b",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:32:53.829299+00:00",
+ "elapsed_seconds": 603.13,
+ "job": "214726",
+ "hostname": "auh7-1b-gpu-312",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d908552fb691/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_t2v.mp4
new file mode 100644
index 0000000..4b1a05f
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d908552fb691/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_editing.jpg
new file mode 100644
index 0000000..18473e4
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_editing.json b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_editing.json
new file mode 100644
index 0000000..dec4942
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "d977ce147d80",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 109,
+ "prompt_truncated": false,
+ "seed": 2026081905,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d977ce147d80/source-exact81.mp4",
+ "source_sha256": "e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "shift_terminal",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "24f7ac9b4d8248228c92995b84173dca81c0bae42f890f507c6fc46f9fae3a43",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:49:26.067248+00:00",
+ "elapsed_seconds": 2309.32,
+ "job": "214748",
+ "hostname": "auh7-1b-gpu-269",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_editing.mp4
new file mode 100644
index 0000000..d7177ae
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_t2v.jpg
new file mode 100644
index 0000000..fbcc542
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_t2v.json
new file mode 100644
index 0000000..7de727b
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_t2v.json
@@ -0,0 +1,104 @@
+{
+ "status": "complete",
+ "case_id": "d977ce147d80",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 70,
+ "prompt_truncated": false,
+ "seed": 2026081905,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "shift_terminal",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "bad7436ab05bc9d57e503b271ad88b897e1a1ee4db3d8c43c940ce83c37cbf1e",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:04:24.410554+00:00",
+ "elapsed_seconds": 578.94,
+ "job": "214703",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_t2v.mp4
new file mode 100644
index 0000000..dc3da37
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d977ce147d80/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_editing.jpg
new file mode 100644
index 0000000..187016e
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_editing.json b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_editing.json
new file mode 100644
index 0000000..8035da6
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "d977ce147d80",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "actual_t5_text": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d977ce147d80",
+ "arm": "editing",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d977ce147d80/source-exact81.mp4",
+ "source_sha256": "e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 83,
+ 2048
+ ],
+ "feature_sha256": "a9f5554f6eb78289b924250b313712b58c925a9b0599cc70399b9a099117334c"
+ },
+ "t5_text_inputs": [
+ "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081905,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d977ce147d80/source-exact81.mp4",
+ "source_sha256": "e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "1ecb531ea6b3a35432fabcb94944940f4a199b8b14067578cf2378b5ab32c23d",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:09:38.268251+00:00",
+ "elapsed_seconds": 1047.39,
+ "job": "214736",
+ "hostname": "auh7-1b-gpu-269",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_editing.mp4
new file mode 100644
index 0000000..30e9016
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_t2v.jpg
new file mode 100644
index 0000000..b3991b8
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_t2v.json b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_t2v.json
new file mode 100644
index 0000000..094c169
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_t2v.json
@@ -0,0 +1,66 @@
+{
+ "status": "complete",
+ "case_id": "d977ce147d80",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "actual_t5_text": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d977ce147d80",
+ "arm": "t2v",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 56,
+ 2048
+ ],
+ "feature_sha256": "f578112393aca285a045449d97541a449320571490c0012b73a7d22c73d52bdd"
+ },
+ "t5_text_inputs": [
+ "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081905,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "802cb826e1a39c7a3e2adcb35e2e2288439bc8acfaa86ec6c68dd450b95601f6",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:21:34.897069+00:00",
+ "elapsed_seconds": 599.87,
+ "job": "214724",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_t2v.mp4
new file mode 100644
index 0000000..7f2f3d3
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/d977ce147d80/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_editing.jpg b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_editing.jpg
new file mode 100644
index 0000000..837e0fc
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_editing.json b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_editing.json
new file mode 100644
index 0000000..7eb3aa7
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_editing.json
@@ -0,0 +1,109 @@
+{
+ "status": "complete",
+ "case_id": "fb7725c351ba",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 107,
+ "prompt_truncated": false,
+ "seed": 2026081901,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/fb7725c351ba/source-exact81.mp4",
+ "source_sha256": "479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "6b544f44db85d07a6f1792758ccb4d4ff2f2482cdf42348661ddf06c33f79575",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:22:14.144770+00:00",
+ "elapsed_seconds": 2187.05,
+ "job": "214692",
+ "hostname": "auh7-1b-gpu-289",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "3b0d2856c6329312cd3627b4df0028cafb7902121d323ef0cdd46ea580f4a200"
+}
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_editing.mp4 b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_editing.mp4
new file mode 100644
index 0000000..3277f2a
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_t2v.jpg b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_t2v.jpg
new file mode 100644
index 0000000..5f79590
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_t2v.json b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_t2v.json
new file mode 100644
index 0000000..df9892d
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_t2v.json
@@ -0,0 +1,77 @@
+{
+ "status": "complete",
+ "case_id": "fb7725c351ba",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "prompt_rewriting": false,
+ "seed": 2026081901,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": null,
+ "expert_usage_assertion_passed": true,
+ "prompt_capture_assertion_passed": true,
+ "full_decode_passed": true,
+ "output": {
+ "sha256": "cb9a41e6af23c63dfcc99430a4c56ba498e97dca5a0f29f3777d6dc47fb5dc32",
+ "width": 832,
+ "height": 480,
+ "avg_frame_rate": "25/1",
+ "nb_read_frames": "81"
+ },
+ "elapsed_seconds": null,
+ "completed_utc": "2026-09-23T19:40:28.952388+00:00",
+ "job": "214649",
+ "hostname": "auh7-1b-gpu-255",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "cf5e73c03b21c57f4ca93a4bfbdd3f6935a35fa3f9e8687736845df1712c49f6",
+ "receipt_recovery": {
+ "reason": "Generated video was valid; unsupported compute-node ffprobe wrapper arguments stopped post-processing.",
+ "observed_failure": "ffprobe subprocess after prompt/expert assertions, encode, and complete ffmpeg decode",
+ "log_path": "logs/bernini-214649.log",
+ "log_sha256": "09c0b4103a2a88d1aaa3c0cb0920acb36ad744a54a0371abc6238423ce9b004a",
+ "driver_path": "provenance/drivers_v1/run_bernini.py",
+ "missing_runtime_measurements": [
+ "per-expert forward counts",
+ "elapsed_seconds",
+ "positive_token_count"
+ ],
+ "validation": "Independent complete local ffmpeg decode and genuine ffprobe frame count; no video regenerated or replaced."
+ },
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "source_reference_frames": [],
+ "prompt_truncated": false,
+ "positive_token_count": null,
+ "positive_token_count_assertion_passed": true,
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ]
+}
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_t2v.mp4 b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_t2v.mp4
new file mode 100644
index 0000000..f297c46
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/fb7725c351ba/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_editing.jpg b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_editing.jpg
new file mode 100644
index 0000000..407bc09
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_editing.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_editing.json b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_editing.json
new file mode 100644
index 0000000..2d1884a
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_editing.json
@@ -0,0 +1,72 @@
+{
+ "status": "complete",
+ "case_id": "fb7725c351ba",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "actual_t5_text": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "fb7725c351ba",
+ "arm": "editing",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/fb7725c351ba/source-exact81.mp4",
+ "source_sha256": "479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 80,
+ 2048
+ ],
+ "feature_sha256": "0f447b68b39d44f6134f90a6a56244778144a651f3c807fce44417faed46405a"
+ },
+ "t5_text_inputs": [
+ "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081901,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/fb7725c351ba/source-exact81.mp4",
+ "source_sha256": "479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "e66930bf62f233054b25761f5b614ac725f044c20ac7179ecca273b7b092493f",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:14:05.087905+00:00",
+ "elapsed_seconds": 932.32,
+ "job": "214720",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+}
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_editing.mp4 b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_editing.mp4
new file mode 100644
index 0000000..63de83e
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_editing.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_t2v.jpg b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_t2v.jpg
new file mode 100644
index 0000000..20dae72
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_t2v.json b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_t2v.json
new file mode 100644
index 0000000..a5db38f
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_t2v.json
@@ -0,0 +1,76 @@
+{
+ "status": "complete",
+ "case_id": "fb7725c351ba",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "prompt_rewriting": false,
+ "seed": 2026081901,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": null,
+ "expert_usage_assertion_passed": true,
+ "prompt_capture_assertion_passed": true,
+ "full_decode_passed": true,
+ "output": {
+ "sha256": "824308c194214741551ae419983cf4c8b2cf0690b1e50d656383ad01d16d86df",
+ "width": 832,
+ "height": 480,
+ "avg_frame_rate": "25/1",
+ "nb_read_frames": "81"
+ },
+ "elapsed_seconds": null,
+ "completed_utc": "2026-09-23T19:53:48.570684+00:00",
+ "job": "214677",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "3e34e9031fb4622896f5ac2765f6c15fcf8740324100c341cf70fb5bb71bba7d",
+ "receipt_recovery": {
+ "reason": "Generated video was valid; unsupported compute-node ffprobe wrapper arguments stopped post-processing.",
+ "observed_failure": "ffprobe subprocess after prompt/expert assertions, encode, and complete ffmpeg decode",
+ "log_path": "logs/omni-canary2-214677.log",
+ "log_sha256": "7de41d5e17edbdb8498006657ad0723392a68896346d72d1856589bbcdd56635",
+ "driver_path": "provenance/drivers_v1/run_omni.py",
+ "missing_runtime_measurements": [
+ "per-expert forward counts",
+ "elapsed_seconds",
+ "negative T5 text not separately persisted; its equality assertion passed"
+ ],
+ "validation": "Independent complete local ffmpeg decode and genuine ffprobe frame count; no video regenerated or replaced."
+ },
+ "actual_t5_text": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "qwen_features": {
+ "case_id": "fb7725c351ba",
+ "arm": "t2v",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 53,
+ 2048
+ ],
+ "feature_sha256": "09257c22c470b91a506cd2f6e96df101e6f138ae2fd62cacc9851d82cd597b7c"
+ },
+ "t5_text_inputs": [
+ "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible."
+ ],
+ "positive_t5_context_copies": 1,
+ "visual_latent_shape": null,
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "total_forward_count_assertion": 80
+}
diff --git a/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_t2v.mp4 b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_t2v.mp4
new file mode 100644
index 0000000..c5161d5
Binary files /dev/null and b/projects/video-action-editing/large-fair12/results/fb7725c351ba/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/large-fair12/summary.json b/projects/video-action-editing/large-fair12/summary.json
new file mode 100644
index 0000000..454e210
--- /dev/null
+++ b/projects/video-action-editing/large-fair12/summary.json
@@ -0,0 +1,4293 @@
+[
+ {
+ "case_id": "fb7725c351ba",
+ "name": "提杠铃 · Barbell lift",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "fb7725c351ba",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "prompt_rewriting": false,
+ "seed": 2026081901,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": null,
+ "expert_usage_assertion_passed": true,
+ "prompt_capture_assertion_passed": true,
+ "full_decode_passed": true,
+ "output": {
+ "sha256": "cb9a41e6af23c63dfcc99430a4c56ba498e97dca5a0f29f3777d6dc47fb5dc32",
+ "width": 832,
+ "height": 480,
+ "avg_frame_rate": "25/1",
+ "nb_read_frames": "81"
+ },
+ "elapsed_seconds": null,
+ "completed_utc": "2026-09-23T19:40:28.952388+00:00",
+ "job": "214649",
+ "hostname": "auh7-1b-gpu-255",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "cf5e73c03b21c57f4ca93a4bfbdd3f6935a35fa3f9e8687736845df1712c49f6",
+ "receipt_recovery": {
+ "reason": "Generated video was valid; unsupported compute-node ffprobe wrapper arguments stopped post-processing.",
+ "observed_failure": "ffprobe subprocess after prompt/expert assertions, encode, and complete ffmpeg decode",
+ "log_path": "logs/bernini-214649.log",
+ "log_sha256": "09c0b4103a2a88d1aaa3c0cb0920acb36ad744a54a0371abc6238423ce9b004a",
+ "driver_path": "provenance/drivers_v1/run_bernini.py",
+ "missing_runtime_measurements": [
+ "per-expert forward counts",
+ "elapsed_seconds",
+ "positive_token_count"
+ ],
+ "validation": "Independent complete local ffmpeg decode and genuine ffprobe frame count; no video regenerated or replaced."
+ },
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "source_reference_frames": [],
+ "prompt_truncated": false,
+ "positive_token_count": null,
+ "positive_token_count_assertion_passed": true,
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ]
+ },
+ {
+ "status": "complete",
+ "case_id": "fb7725c351ba",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "prompt_rewriting": false,
+ "seed": 2026081901,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": null,
+ "expert_usage_assertion_passed": true,
+ "prompt_capture_assertion_passed": true,
+ "full_decode_passed": true,
+ "output": {
+ "sha256": "824308c194214741551ae419983cf4c8b2cf0690b1e50d656383ad01d16d86df",
+ "width": 832,
+ "height": 480,
+ "avg_frame_rate": "25/1",
+ "nb_read_frames": "81"
+ },
+ "elapsed_seconds": null,
+ "completed_utc": "2026-09-23T19:53:48.570684+00:00",
+ "job": "214677",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "3e34e9031fb4622896f5ac2765f6c15fcf8740324100c341cf70fb5bb71bba7d",
+ "receipt_recovery": {
+ "reason": "Generated video was valid; unsupported compute-node ffprobe wrapper arguments stopped post-processing.",
+ "observed_failure": "ffprobe subprocess after prompt/expert assertions, encode, and complete ffmpeg decode",
+ "log_path": "logs/omni-canary2-214677.log",
+ "log_sha256": "7de41d5e17edbdb8498006657ad0723392a68896346d72d1856589bbcdd56635",
+ "driver_path": "provenance/drivers_v1/run_omni.py",
+ "missing_runtime_measurements": [
+ "per-expert forward counts",
+ "elapsed_seconds",
+ "negative T5 text not separately persisted; its equality assertion passed"
+ ],
+ "validation": "Independent complete local ffmpeg decode and genuine ffprobe frame count; no video regenerated or replaced."
+ },
+ "actual_t5_text": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "qwen_features": {
+ "case_id": "fb7725c351ba",
+ "arm": "t2v",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 53,
+ 2048
+ ],
+ "feature_sha256": "09257c22c470b91a506cd2f6e96df101e6f138ae2fd62cacc9851d82cd597b7c"
+ },
+ "t5_text_inputs": [
+ "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible."
+ ],
+ "positive_t5_context_copies": 1,
+ "visual_latent_shape": null,
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "total_forward_count_assertion": 80
+ },
+ {
+ "status": "complete",
+ "case_id": "fb7725c351ba",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 107,
+ "prompt_truncated": false,
+ "seed": 2026081901,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/fb7725c351ba/source-exact81.mp4",
+ "source_sha256": "479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "6b544f44db85d07a6f1792758ccb4d4ff2f2482cdf42348661ddf06c33f79575",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:22:14.144770+00:00",
+ "elapsed_seconds": 2187.05,
+ "job": "214692",
+ "hostname": "auh7-1b-gpu-289",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "3b0d2856c6329312cd3627b4df0028cafb7902121d323ef0cdd46ea580f4a200"
+ },
+ {
+ "status": "complete",
+ "case_id": "fb7725c351ba",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "actual_t5_text": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "fb7725c351ba",
+ "arm": "editing",
+ "caption": "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "caption_sha256": "b3337cc9b01e086c7969770d68fc6c6572e454e5fe1be64f770a6b7e1e30500d",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/fb7725c351ba/source-exact81.mp4",
+ "source_sha256": "479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 80,
+ 2048
+ ],
+ "feature_sha256": "0f447b68b39d44f6134f90a6a56244778144a651f3c807fce44417faed46405a"
+ },
+ "t5_text_inputs": [
+ "The woman in a home gym lifts a barbell from the floor to her waist. She wears black activewear. The home gym has a tiled floor and exercise equipment. A static full-body shot keeps the woman and the barbell visible.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081901,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/fb7725c351ba/source-exact81.mp4",
+ "source_sha256": "479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "e66930bf62f233054b25761f5b614ac725f044c20ac7179ecca273b7b092493f",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:14:05.087905+00:00",
+ "elapsed_seconds": 932.32,
+ "job": "214720",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "3fadbe0c2961",
+ "name": "狗转头 · Dog head turn",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "3fadbe0c2961",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 63,
+ "prompt_truncated": false,
+ "seed": 2026081902,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "dc894ddffe1960fa0a845b2bbf03c37162d8b496e887ef535bed5cbf2ace9616",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:45:29.419504+00:00",
+ "elapsed_seconds": 706.21,
+ "job": "214706",
+ "hostname": "auh7-1b-gpu-269",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "3fadbe0c2961",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "actual_t5_text": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "3fadbe0c2961",
+ "arm": "t2v",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 51,
+ 2048
+ ],
+ "feature_sha256": "13a1edc8e7db94239ea6c6049d5691001949d9c95c3900fb5bbfd9298401f18a"
+ },
+ "t5_text_inputs": [
+ "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081902,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "e92d7a1bcddf3582e33dbd1b93244e9a12de37a18ea27ffff2088f026e3584ec",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:35:53.507686+00:00",
+ "elapsed_seconds": 637.05,
+ "job": "214727",
+ "hostname": "auh7-1b-gpu-307",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "3fadbe0c2961",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 102,
+ "prompt_truncated": false,
+ "seed": 2026081902,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/3fadbe0c2961/source-exact81.mp4",
+ "source_sha256": "99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "sigma_max",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "59ddf05fd48e62dffbc27aac5a721cb7198c40d526c06479056695547d9cf48f",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:49:23.038345+00:00",
+ "elapsed_seconds": 1621.77,
+ "job": "214751",
+ "hostname": "auh7-1b-gpu-307",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "3fadbe0c2961",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "actual_t5_text": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "3fadbe0c2961",
+ "arm": "editing",
+ "caption": "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "caption_sha256": "aaeb58cbd788e3a815962e6289b1853ca09f2ef13c69d98a1af484e67f9da178",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/3fadbe0c2961/source-exact81.mp4",
+ "source_sha256": "99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 78,
+ 2048
+ ],
+ "feature_sha256": "5607250394671775f7a81e3929b7e9ff08652121e7def64d0a370240e1f6e2ce"
+ },
+ "t5_text_inputs": [
+ "The dog turns its head further to the right while sitting in the grass. The dog is a reddish-brown and white Pembroke Welsh Corgi in a sunny green field. The camera is stationary in a medium shot.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081902,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/3fadbe0c2961/source-exact81.mp4",
+ "source_sha256": "99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "87aa2a665e5c451327950cf7eaa217e0f10ba00a2dac3f85af22489eb046501b",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:07:32.501671+00:00",
+ "elapsed_seconds": 637.79,
+ "job": "214739",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "7077cfb160cf",
+ "name": "长椅上写字 · Write in notebook",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "7077cfb160cf",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 68,
+ "prompt_truncated": false,
+ "seed": 2026081904,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "shift_terminal",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "eaf121f7e20fa7dc7aa6d4d127b00180cb92fee67e1fdea5ff2fbc7b9be7d997",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:19:58.997189+00:00",
+ "elapsed_seconds": 705.61,
+ "job": "214757",
+ "hostname": "auh7-1b-gpu-269",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "7077cfb160cf",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "actual_t5_text": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7077cfb160cf",
+ "arm": "t2v",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 53,
+ 2048
+ ],
+ "feature_sha256": "863e64e016115388d775caa7fd1a8be61270b967361152418dfda43d47ba5cc6"
+ },
+ "t5_text_inputs": [
+ "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081904,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "cfce2698573f7b20a42686a08059ff6d39d7053cafcf4731fddab4d174c804ae",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:35:38.009473+00:00",
+ "elapsed_seconds": 599.71,
+ "job": "214728",
+ "hostname": "auh7-1b-gpu-283",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "7077cfb160cf",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 107,
+ "prompt_truncated": false,
+ "seed": 2026081904,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7077cfb160cf/source-exact81.mp4",
+ "source_sha256": "cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "sigma_min",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "2b0371a6e0660520daf895823f2f74b556503b075578fe2dca02f5c398bd23e7",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:05:32.304246+00:00",
+ "elapsed_seconds": 2298.97,
+ "job": "214752",
+ "hostname": "auh7-1b-gpu-210",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "7077cfb160cf",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "actual_t5_text": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7077cfb160cf",
+ "arm": "editing",
+ "caption": "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "caption_sha256": "0acdffedc18b04d2c2bef7d0a4e4e51544b75d4dd1f904e080d9e0490be24fb1",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7077cfb160cf/source-exact81.mp4",
+ "source_sha256": "cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 80,
+ 2048
+ ],
+ "feature_sha256": "70de5d8bbce06459f31e50438e02db4aae638a7c14eec75b64875ee09eb289cb"
+ },
+ "t5_text_inputs": [
+ "The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. She has shoulder-length light brown hair and wears a pink shirt and blue jeans. The wooden bench is in a sunny park. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081904,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7077cfb160cf/source-exact81.mp4",
+ "source_sha256": "cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "3b158d997abb43e583b01b618a280bb0938570e81b73116ce351f2cb255e96af",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:08:05.180195+00:00",
+ "elapsed_seconds": 649.79,
+ "job": "214740",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "7e488525099e",
+ "name": "转身倚车 · Turn and lean on car",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "7e488525099e",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 58,
+ "prompt_truncated": false,
+ "seed": 2026081906,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "sigma_min",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "4d3b693a341087b216eba42bc9933ed61efbead441536c78638dfd4e5cca9145",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:17:05.780853+00:00",
+ "elapsed_seconds": 519.14,
+ "job": "214758",
+ "hostname": "auh7-1b-gpu-312",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "7e488525099e",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "actual_t5_text": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7e488525099e",
+ "arm": "t2v",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 44,
+ 2048
+ ],
+ "feature_sha256": "9f6d9a9afe61984fcb7355e65810471e2b4b623b73134d18a47daa0936b86b83"
+ },
+ "t5_text_inputs": [
+ "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081906,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "5d79f8d88fc14bccd0336603594e03abcc72413534b835b37698aed74b1bc7cf",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:36:39.023949+00:00",
+ "elapsed_seconds": 615.52,
+ "job": "214729",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "7e488525099e",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 97,
+ "prompt_truncated": false,
+ "seed": 2026081906,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7e488525099e/source-exact81.mp4",
+ "source_sha256": "1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "9829080372e3351bb3375d787615f72cc4a10be693f382edd9d36fae37079d0e",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:38:14.312525+00:00",
+ "elapsed_seconds": 2268.32,
+ "job": "214753",
+ "hostname": "auh7-1b-gpu-221",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "7e488525099e",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "actual_t5_text": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7e488525099e",
+ "arm": "editing",
+ "caption": "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "caption_sha256": "ed6115b291eb39710054250e5a6fd3e51f92ed3675fe1af1c5b45617396c3082",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7e488525099e/source-exact81.mp4",
+ "source_sha256": "1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 71,
+ 2048
+ ],
+ "feature_sha256": "a34cbe81e1903e91618292d5ab323f49478f88c4638de853db8d3bb4475d4242"
+ },
+ "t5_text_inputs": [
+ "The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The setting is an underground parking garage with steady red lighting. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081906,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7e488525099e/source-exact81.mp4",
+ "source_sha256": "1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "19321c8236ea55c3ce3006817c0c8316f9313bc193729e4ba6dd1651f445d5fc",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:15:57.054117+00:00",
+ "elapsed_seconds": 886.52,
+ "job": "214741",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "d5221a1b69d5",
+ "name": "闭眼尝汤 · Taste from ladle",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "d5221a1b69d5",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 53,
+ "prompt_truncated": false,
+ "seed": 2026081907,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "28120017a5e122b3f72956b9c2b16b20aae762baa4e88b9bab605ed03fff7c40",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:42:17.028942+00:00",
+ "elapsed_seconds": 483.92,
+ "job": "214709",
+ "hostname": "auh7-1b-gpu-312",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "d5221a1b69d5",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "actual_t5_text": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d5221a1b69d5",
+ "arm": "t2v",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 40,
+ 2048
+ ],
+ "feature_sha256": "32fb5857a4b6f85b7d8f9028089619067c830e83ddb787da525a2475cfdcbf7c"
+ },
+ "t5_text_inputs": [
+ "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081907,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "9ee4d254b5c2d579d3a9adcdc9d1057d7a8ef0afe1860f8a785db66d2cedd8ab",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:36:30.235433+00:00",
+ "elapsed_seconds": 603.1,
+ "job": "214730",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "d5221a1b69d5",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 92,
+ "prompt_truncated": false,
+ "seed": 2026081907,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d5221a1b69d5/source-exact81.mp4",
+ "source_sha256": "fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "shift_terminal",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "9cdbd07566d899630c7d460876abe733604cb04d284adf01842a9716a5db4ec3",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:38:07.139056+00:00",
+ "elapsed_seconds": 2248.64,
+ "job": "214754",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "d5221a1b69d5",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "actual_t5_text": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d5221a1b69d5",
+ "arm": "editing",
+ "caption": "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "caption_sha256": "043306a7c3eff9ad7a50db5bbf0d3b4a2d6dec27fb8c5b5f6a316382992bc71b",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d5221a1b69d5/source-exact81.mp4",
+ "source_sha256": "fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 67,
+ 2048
+ ],
+ "feature_sha256": "c34f3a4e8b0058cf7a7f941082a5266c00ee63aea17fccbbdec08c7d9905eddb"
+ },
+ "t5_text_inputs": [
+ "The woman in an apron closes her eyes as she tastes from a metal ladle. She wears a patterned apron in a modern kitchen. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081907,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d5221a1b69d5/source-exact81.mp4",
+ "source_sha256": "fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "a1e43492995e5fefe7ace547ce6f62cb6fe363ae29ad365f7144f36f7bf4e900",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:23:04.739246+00:00",
+ "elapsed_seconds": 963.54,
+ "job": "214742",
+ "hostname": "auh7-1b-gpu-210",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "40712e1341dc",
+ "name": "放下手机后翻身睡觉 · Put phone away and sleep",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "40712e1341dc",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 74,
+ "prompt_truncated": false,
+ "seed": 2026081911,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "sigma_max",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "ac158010a45d2c746774740d9a513440d29ac8d82604895d75b41c73e1461eef",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:45:13.322144+00:00",
+ "elapsed_seconds": 482.84,
+ "job": "214710",
+ "hostname": "auh7-1b-gpu-283",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "40712e1341dc",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "actual_t5_text": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "40712e1341dc",
+ "arm": "t2v",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 61,
+ 2048
+ ],
+ "feature_sha256": "f953db0ef97435bed85eb46eb0a25136e1775f7d4342e1a435a9ed189580145c"
+ },
+ "t5_text_inputs": [
+ "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081911,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "9dafe547fd5b649dccbf097319e9f41141518721416a7a75534548dbf4f138e3",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:26:43.210982+00:00",
+ "elapsed_seconds": 958.87,
+ "job": "214756",
+ "hostname": "auh7-1b-gpu-210",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "40712e1341dc",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 113,
+ "prompt_truncated": false,
+ "seed": 2026081911,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/40712e1341dc/source-exact81.mp4",
+ "source_sha256": "654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_max",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "d6b8387a8f977f88fec1fa6f16ef0ad0ae9c22554a74bb661014ad6260532037",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T22:30:33.960692+00:00",
+ "elapsed_seconds": 1605.68,
+ "job": "214755",
+ "hostname": "auh7-1b-gpu-283",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "40712e1341dc",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "actual_t5_text": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "40712e1341dc",
+ "arm": "editing",
+ "caption": "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "caption_sha256": "24857d19e619899acd653ea104c6a5b68b895cdd4b5cc3637df661f694521907",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/40712e1341dc/source-exact81.mp4",
+ "source_sha256": "654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 88,
+ 2048
+ ],
+ "feature_sha256": "17f26bdae837dadcbf1c11b139c2c9ba8cbaa805f2f35c4fff16b1a5a0a45f67"
+ },
+ "t5_text_inputs": [
+ "The man in bed puts his smartphone on the nightstand and turns over to go to sleep. He has dyed-blond hair, wears a pink tank top and a nose ring, and is in a dark bedroom with a nightstand beside the bed. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081911,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/40712e1341dc/source-exact81.mp4",
+ "source_sha256": "654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "271248625a365fabc0664f1e5f3cfadd94ad2179e17a2efa261b9d5926d3765b",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:20:56.467352+00:00",
+ "elapsed_seconds": 646.24,
+ "job": "214743",
+ "hostname": "auh7-1b-gpu-307",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "10af6c651dbc",
+ "name": "Jellyfish contraction",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "10af6c651dbc",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 65,
+ "prompt_truncated": false,
+ "seed": 2026081910,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "99c080cb8ebc719e47354dbc79ba641c7ffcc30c355bbafd05335903314ba8d8",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:04:25.056432+00:00",
+ "elapsed_seconds": 579.95,
+ "job": "214700",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "10af6c651dbc",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "actual_t5_text": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "10af6c651dbc",
+ "arm": "t2v",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 47,
+ 2048
+ ],
+ "feature_sha256": "b1d9193dbd160064ceb99196242d2707a2f74e510993302982dac03e6b713c01"
+ },
+ "t5_text_inputs": [
+ "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081910,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "1d1edd0eef41ff1fcde1a1bd662fdb47694a3b9ed60e458e83e842af441f9eb6",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:18:02.505227+00:00",
+ "elapsed_seconds": 608.46,
+ "job": "214721",
+ "hostname": "auh7-1b-gpu-312",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "10af6c651dbc",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 104,
+ "prompt_truncated": false,
+ "seed": 2026081910,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/10af6c651dbc/source-exact81.mp4",
+ "source_sha256": "ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "sigma_min",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "82c8810cd7546c2d6ea80743c3c90d4a926e37f5099790e2b6acf5fc19c77510",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:36:06.353806+00:00",
+ "elapsed_seconds": 1623.09,
+ "job": "214745",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "10af6c651dbc",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "actual_t5_text": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "10af6c651dbc",
+ "arm": "editing",
+ "caption": "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "caption_sha256": "790782c54c3fea6d88d28f9129501ef060ebc863cd6ead101f7a12563d43e383",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/10af6c651dbc/source-exact81.mp4",
+ "source_sha256": "ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 74,
+ 2048
+ ],
+ "feature_sha256": "31d5a8e438020fe75f1d26e7ad419c5e28ba4983c9e49203dd1a52efd9017d75"
+ },
+ "t5_text_inputs": [
+ "The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. A small, translucent jellyfish with white spots is surrounded by deep blue water. The underwater camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081910,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/10af6c651dbc/source-exact81.mp4",
+ "source_sha256": "ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "d7a3b42e059c8ac6133f96b2fc4664509d504c9f821f45cc24d9d99acb151d44",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:56:21.408682+00:00",
+ "elapsed_seconds": 921.01,
+ "job": "214733",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "7fa338474c75",
+ "name": "Playful selfie",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "7fa338474c75",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 61,
+ "prompt_truncated": false,
+ "seed": 2026081900,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_max",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "9f68a732503e5d11d4b94ee86ea61175d555092ffc539b69615faf0e050a083d",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:03:56.015342+00:00",
+ "elapsed_seconds": 558.77,
+ "job": "214701",
+ "hostname": "auh7-1b-gpu-283",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "7fa338474c75",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "actual_t5_text": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7fa338474c75",
+ "arm": "t2v",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 46,
+ 2048
+ ],
+ "feature_sha256": "d54fddcf16850d14dc9dfb71b863ba619646a35c90cd4dd2d62419a2adda477a"
+ },
+ "t5_text_inputs": [
+ "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081900,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "0c6b53c4bae5fa1c7275fb71f5197c6312876d3813ae35b9c4db361712d3c65f",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:20:46.292401+00:00",
+ "elapsed_seconds": 595.85,
+ "job": "214722",
+ "hostname": "auh7-1b-gpu-283",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "7fa338474c75",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 100,
+ "prompt_truncated": false,
+ "seed": 2026081900,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7fa338474c75/source-exact81.mp4",
+ "source_sha256": "4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "sigma_max",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "d80fbfb74a6ab3aed91b258c759a2eb21e37b66c82513742590b0e512e0f49f4",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:36:39.558813+00:00",
+ "elapsed_seconds": 1623.02,
+ "job": "214746",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "7fa338474c75",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "actual_t5_text": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "7fa338474c75",
+ "arm": "editing",
+ "caption": "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "caption_sha256": "f7d22ce2d98ed6adff42d53192e7204016d9035165349de21317c756823ac96e",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7fa338474c75/source-exact81.mp4",
+ "source_sha256": "4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 73,
+ 2048
+ ],
+ "feature_sha256": "dc6a8fcf6aace395826a95d45c233eabfd1bbacf9f833e8de912e13129a059d8"
+ },
+ "t5_text_inputs": [
+ "The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. She is framed in a static medium shot against a uniform green screen background.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081900,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7fa338474c75/source-exact81.mp4",
+ "source_sha256": "4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "e4267370c4b549bba46b2c14020f9fff2a725e1b4c90379041ab73bbeb59e8be",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:51:53.409740+00:00",
+ "elapsed_seconds": 636.88,
+ "job": "214734",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "81533c9e56ec",
+ "name": "Scarf spin",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "81533c9e56ec",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 65,
+ "prompt_truncated": false,
+ "seed": 2026081915,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "sigma_max",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "d7a3869e2f1e3f30b90495f183234d1dcd56b5566c27678ac94669f243f45be8",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:03:06.079836+00:00",
+ "elapsed_seconds": 499.76,
+ "job": "214702",
+ "hostname": "auh7-1b-gpu-312",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "81533c9e56ec",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "actual_t5_text": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "81533c9e56ec",
+ "arm": "t2v",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 49,
+ 2048
+ ],
+ "feature_sha256": "55500ce71d10d808501849ae5c122f61dbff99b38bd16dd415fb146d599a62c3"
+ },
+ "t5_text_inputs": [
+ "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081915,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "a78f02ecb3aaa875cec8a133397227a526d4ea0e8dadf0a79bb3d7f8478f49e2",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:21:31.826216+00:00",
+ "elapsed_seconds": 612.01,
+ "job": "214723",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "81533c9e56ec",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 104,
+ "prompt_truncated": false,
+ "seed": 2026081915,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/81533c9e56ec/source-exact81.mp4",
+ "source_sha256": "6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "shift_terminal",
+ "sigma_min",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "6568610950f2753ca3d4c6812018131d53241cb7921236502d6b4d4946380dd9",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:47:28.810955+00:00",
+ "elapsed_seconds": 2245.25,
+ "job": "214747",
+ "hostname": "auh7-1b-gpu-206",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "81533c9e56ec",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "actual_t5_text": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "81533c9e56ec",
+ "arm": "editing",
+ "caption": "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "caption_sha256": "f6e77c2c4c9a431ed92369c65484776cb036d47a4d72c6a19e11d2a4d5b09579",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/81533c9e56ec/source-exact81.mp4",
+ "source_sha256": "6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 76,
+ 2048
+ ],
+ "feature_sha256": "09cbf7460835b917d1811c9612fb73eb0bd7c7ad17b988a8093c86702c180100"
+ },
+ "t5_text_inputs": [
+ "The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. Her scarf is red and white. A flower-lined fence and a rustic wooden house form the background. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081915,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/81533c9e56ec/source-exact81.mp4",
+ "source_sha256": "6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "d156e96a7fb7d91ea4a5e0159841362af432c9fed0b20882a8d508796398f671",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:52:22.199956+00:00",
+ "elapsed_seconds": 650.56,
+ "job": "214735",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "d977ce147d80",
+ "name": "Rub eyes then cover face",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "d977ce147d80",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 70,
+ "prompt_truncated": false,
+ "seed": 2026081905,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "shift_terminal",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "bad7436ab05bc9d57e503b271ad88b897e1a1ee4db3d8c43c940ce83c37cbf1e",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:04:24.410554+00:00",
+ "elapsed_seconds": 578.94,
+ "job": "214703",
+ "hostname": "auh7-1b-gpu-313",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "d977ce147d80",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "actual_t5_text": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d977ce147d80",
+ "arm": "t2v",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 56,
+ 2048
+ ],
+ "feature_sha256": "f578112393aca285a045449d97541a449320571490c0012b73a7d22c73d52bdd"
+ },
+ "t5_text_inputs": [
+ "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081905,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "802cb826e1a39c7a3e2adcb35e2e2288439bc8acfaa86ec6c68dd450b95601f6",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:21:34.897069+00:00",
+ "elapsed_seconds": 599.87,
+ "job": "214724",
+ "hostname": "auh7-1b-gpu-282",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "d977ce147d80",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 109,
+ "prompt_truncated": false,
+ "seed": 2026081905,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d977ce147d80/source-exact81.mp4",
+ "source_sha256": "e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "shift_terminal",
+ "sigma_max"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "24f7ac9b4d8248228c92995b84173dca81c0bae42f890f507c6fc46f9fae3a43",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:49:26.067248+00:00",
+ "elapsed_seconds": 2309.32,
+ "job": "214748",
+ "hostname": "auh7-1b-gpu-269",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "d977ce147d80",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "actual_t5_text": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d977ce147d80",
+ "arm": "editing",
+ "caption": "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "caption_sha256": "a217f2a0d32d07e58d5c5781d12d401ed7a08f25d9286ea9f0bee5c25677e960",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d977ce147d80/source-exact81.mp4",
+ "source_sha256": "e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 83,
+ 2048
+ ],
+ "feature_sha256": "a9f5554f6eb78289b924250b313712b58c925a9b0599cc70399b9a099117334c"
+ },
+ "t5_text_inputs": [
+ "The man lying on the couch rubs his tired eyes and covers his face with both hands. He has dark skin, dreadlocks, and a gold septum ring. The couch is dark and the room is dimly lit. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081905,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d977ce147d80/source-exact81.mp4",
+ "source_sha256": "e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "1ecb531ea6b3a35432fabcb94944940f4a199b8b14067578cf2378b5ab32c23d",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:09:38.268251+00:00",
+ "elapsed_seconds": 1047.39,
+ "job": "214736",
+ "hostname": "auh7-1b-gpu-269",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "840b214afead",
+ "name": "Place bottle on treadmill",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "840b214afead",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 71,
+ "prompt_truncated": false,
+ "seed": 2026081913,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "sigma_min",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "88c0e40f239232c6cbb95be01496886f9100f26023b357bc13dd26715e4155cb",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:49:44.701325+00:00",
+ "elapsed_seconds": 545.02,
+ "job": "214704",
+ "hostname": "auh7-1b-gpu-307",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "840b214afead",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "actual_t5_text": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "840b214afead",
+ "arm": "t2v",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 54,
+ 2048
+ ],
+ "feature_sha256": "dd48866a02a17042eef8b6cbde060576c10deaf6137ee965388861edfd103d7a"
+ },
+ "t5_text_inputs": [
+ "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081913,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "86baea5251e3c110b691d72b25cfb9405e3039affbbaf8e921c7e9169101b1c2",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:36:13.591878+00:00",
+ "elapsed_seconds": 1029.69,
+ "job": "214725",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "840b214afead",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 110,
+ "prompt_truncated": false,
+ "seed": 2026081913,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/840b214afead/source-exact81.mp4",
+ "source_sha256": "a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_min",
+ "sigma_max",
+ "shift_terminal"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "aa4f8096577b9224c04eadfefb1581272b7a1355a9a730939813296ef15d39a2",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:59:20.369231+00:00",
+ "elapsed_seconds": 2401.86,
+ "job": "214749",
+ "hostname": "auh7-1b-gpu-320",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "840b214afead",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "actual_t5_text": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "840b214afead",
+ "arm": "editing",
+ "caption": "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "caption_sha256": "fe3b668a66485813f97192ae443c497bc34efcc6edbd10bbdf8af1724af85247",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/840b214afead/source-exact81.mp4",
+ "source_sha256": "a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 81,
+ 2048
+ ],
+ "feature_sha256": "8d2079d5bd9b6692dfeb7a739b208a4eaf19cf1aca8b431b9a99ec76318349d8"
+ },
+ "t5_text_inputs": [
+ "The woman turns her head to the left and places her water bottle onto the treadmill. She has long reddish-blonde hair and wears a mint green athletic outfit. The modern gym has blue and white striped pillars. The camera is stationary.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081913,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/840b214afead/source-exact81.mp4",
+ "source_sha256": "a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "8e7370fc84142be3daa707d6ffc7ff82556f7167a6fe14ed26ce5112d574f625",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:08:46.895702+00:00",
+ "elapsed_seconds": 974.76,
+ "job": "214737",
+ "hostname": "auh7-1b-gpu-206",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ },
+ {
+ "case_id": "d908552fb691",
+ "name": "Bow to clap",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "outputs": [
+ {
+ "status": "complete",
+ "case_id": "d908552fb691",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "t2v",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "actual_t5_text": "You are a helpful assistant specialized in text-to-video generation.The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "task_prefix": "You are a helpful assistant specialized in text-to-video generation.",
+ "native_reference_binding": "",
+ "prompt_rewriting": false,
+ "positive_token_count": 60,
+ "prompt_truncated": false,
+ "seed": 2026081903,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "shift_terminal",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 26,
+ 54
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "4316ede62030399da548d0b2b136333c622705fe1395b948ab7d60eca68430ce",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:45:46.786299+00:00",
+ "elapsed_seconds": 730.83,
+ "job": "214705",
+ "hostname": "auh7-1b-gpu-206",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "d908552fb691",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "t2v",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "actual_t5_text": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d908552fb691",
+ "arm": "t2v",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "source_video_path": null,
+ "source_sha256": null,
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 51,
+ 2048
+ ],
+ "feature_sha256": "104ba746d40a2909261000fd07d3cefaf8f075beea1e2f6d4224a62c95937b01"
+ },
+ "t5_text_inputs": [
+ "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 1,
+ "seed": 2026081903,
+ "steps": 40,
+ "fps": 25,
+ "source_video": null,
+ "source_sha256": null,
+ "real_target_used": false,
+ "visual_latent_shape": null,
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "7458de291881bf7e8eb7909a7dbad8fae1c01726304a6494eb637e4ba845c60b",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T20:32:53.829299+00:00",
+ "elapsed_seconds": 603.13,
+ "job": "214726",
+ "hostname": "auh7-1b-gpu-312",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ },
+ {
+ "status": "complete",
+ "case_id": "d908552fb691",
+ "model": "ByteDance/Bernini-R-Diffusers",
+ "revision": "de8c4621d3ac75cc33efe3db8deaed2023e9ac8c",
+ "arm": "editing",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "actual_t5_text": "You are a helpful assistant specialized in video editing with reference.image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "task_prefix": "You are a helpful assistant specialized in video editing with reference.",
+ "native_reference_binding": "image0, image1, image2, and image3 are frames sampled from the source video. Preserve the source subject's identity, background, and camera while applying the requested new action. ",
+ "prompt_rewriting": false,
+ "positive_token_count": 99,
+ "prompt_truncated": false,
+ "seed": 2026081903,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d908552fb691/source-exact81.mp4",
+ "source_sha256": "56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb",
+ "effective_scheduler_config": {
+ "num_train_timesteps": 1000,
+ "beta_start": 0.0001,
+ "beta_end": 0.02,
+ "beta_schedule": "linear",
+ "trained_betas": null,
+ "solver_order": 2,
+ "prediction_type": "flow_prediction",
+ "thresholding": false,
+ "dynamic_thresholding_ratio": 0.995,
+ "sample_max_value": 1.0,
+ "predict_x0": true,
+ "solver_type": "bh2",
+ "lower_order_final": true,
+ "disable_corrector": [],
+ "solver_p": null,
+ "use_karras_sigmas": false,
+ "use_exponential_sigmas": false,
+ "use_beta_sigmas": false,
+ "use_flow_sigmas": true,
+ "flow_shift": 3.0,
+ "timestep_spacing": "linspace",
+ "steps_offset": 0,
+ "final_sigmas_type": "zero",
+ "rescale_betas_zero_snr": false,
+ "use_dynamic_shifting": false,
+ "time_shift_type": "exponential",
+ "sigma_min": null,
+ "sigma_max": null,
+ "shift_terminal": null,
+ "_use_default_values": [
+ "sigma_max",
+ "shift_terminal",
+ "sigma_min"
+ ],
+ "_class_name": "UniPCMultistepScheduler",
+ "_diffusers_version": "0.35.0.dev0"
+ },
+ "source_reference_frames": [
+ 0,
+ 27,
+ 53,
+ 80
+ ],
+ "real_target_used": false,
+ "expert_parameters": [
+ 14288491584,
+ 14288491584
+ ],
+ "expert_forward_counts": [
+ 52,
+ 108
+ ],
+ "loading_info": [
+ {
+ "subfolder": "transformer",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ },
+ {
+ "subfolder": "transformer_2",
+ "info": {
+ "missing_keys": [],
+ "unexpected_keys": [],
+ "mismatched_keys": [],
+ "error_msgs": []
+ }
+ }
+ ],
+ "output": {
+ "sha256": "f164b04e01c335a49aca4f94494b87bd5ff2253a61f7d158fa2ddd00ade9967b",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 464,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:59:03.774298+00:00",
+ "elapsed_seconds": 2250.01,
+ "job": "214750",
+ "hostname": "auh7-1b-gpu-221",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "700a4e76177713a6ba2ff7bb92d7976788b646bf153ad7154701ed422ca966d6"
+ },
+ {
+ "status": "complete",
+ "case_id": "d908552fb691",
+ "model": "Fudan-FUXI/OmniVideo2-A14B",
+ "revision": "51160486a5498cce099bde02cbf232cd046ab4a9",
+ "arm": "editing",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "actual_t5_text": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "prompt_rewriting": false,
+ "qwen_features": {
+ "case_id": "d908552fb691",
+ "arm": "editing",
+ "caption": "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "caption_sha256": "025f04cb7e601915763666a56d60e01f88ed30d6efa85a172c0cf38eff834896",
+ "source_video_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d908552fb691/source-exact81.mp4",
+ "source_sha256": "56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb",
+ "rewriting_enabled": false,
+ "autoregressive_generate_forbidden": true,
+ "qwen_checkpoint": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "feature_shape": [
+ 78,
+ 2048
+ ],
+ "feature_sha256": "114740c8405aa14069531f4ff749cc0da9706f5fe07cf087a02f9e11bbdbc19c"
+ },
+ "t5_text_inputs": [
+ "The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. Her uniform is white with a black collar. The background is black with side lighting, in a static medium shot.",
+ "overly vivid colors, overexposed, static, blurry and unclear details, subtitles, text overlays, stylized artwork, painting, illustration, still image, frozen frame, grayish overall tone, worst quality, low quality, JPEG compression artifacts, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn face, deformed, disfigured, malformed anatomy, distorted limbs, fused fingers, motionless frame, cluttered background, three legs, walking backward, expression drift, teeth flickering, uneven facial muscles, facial shivering"
+ ],
+ "positive_t5_context_copies": 2,
+ "seed": 2026081903,
+ "steps": 40,
+ "fps": 25,
+ "source_video": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d908552fb691/source-exact81.mp4",
+ "source_sha256": "56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb",
+ "real_target_used": false,
+ "visual_latent_shape": [
+ 1,
+ 16,
+ 21,
+ 60,
+ 104
+ ],
+ "expert_parameters": [
+ 14319177792,
+ 14319177792
+ ],
+ "expert_forward_counts": [
+ 34,
+ 46
+ ],
+ "strict_checkpoint_load": true,
+ "max_context_length": 9216,
+ "context_truncation": false,
+ "output": {
+ "sha256": "aafdce633f1ac4a24120c74a0a75ddb3a4b8a4198e0734ce3643bf8124d8ceb2",
+ "avg_frame_rate": "25/1",
+ "codec_name": "h264",
+ "duration": "3.240000000",
+ "height": 480,
+ "nb_read_frames": "81",
+ "width": 832
+ },
+ "full_decode_passed": true,
+ "completed_utc": "2026-09-23T21:05:24.214802+00:00",
+ "elapsed_seconds": 651.09,
+ "job": "214738",
+ "hostname": "auh7-1b-gpu-307",
+ "inputs_sha256": "caa915bc2243c0e21f38532c0b4100e7e4078e575d5d14f54656edc91cd3a535",
+ "driver_sha256": "c83cf03c774b3f29f430e9442d6667e5d7d26e7813a836379620c875e83b8295"
+ }
+ ]
+ }
+]
diff --git a/projects/video-action-editing/metadata.yaml b/projects/video-action-editing/metadata.yaml
new file mode 100644
index 0000000..ac48a7c
--- /dev/null
+++ b/projects/video-action-editing/metadata.yaml
@@ -0,0 +1,14 @@
+title: "Video Action Editing"
+slug: video-action-editing
+status: active
+summary: "VADD: Video Action Editing via Demonstrations and Disentanglement, with theory, methods, and interactive comparisons on synthetic and real videos."
+leaders:
+ - name: Peiyuan Zhu
+ short_name: peiyuan
+ github: SoftPointer
+tags:
+ - video-editing
+ - action-editing
+ - disentanglement
+links:
+ demo: https://causalvideogroup.github.io/projects/video-action-editing/
diff --git a/projects/video-action-editing/omni-all12/index.html b/projects/video-action-editing/omni-all12/index.html
new file mode 100644
index 0000000..230fe6d
--- /dev/null
+++ b/projects/video-action-editing/omni-all12/index.html
@@ -0,0 +1,23 @@
+
+Bernini / OmniVideo2 · T2V 与视频编辑十二例对比
文本核对：Omni 扩写把 lifts from the floor to her waist 写成 lowering ... from the floor to her waist，动作方向语义矛盾。初始整体 caption 还描述重复 deadlifts。
Caption 来源与实际生成文本（展开对照）
目标动作标注 · 与 Editing instruction 的动作正文对应
The woman in a home gym lifts a barbell from the floor to her waist.
原始整体描述 · MEV global_prompt.short_caption
A woman in black activewear performs barbell deadlifts with focused determination in a well-equipped home gym.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A woman in black activewear performs barbell deadlifts with focused determination in a well-equipped home gym. In this continuous shot, The woman in a home gym lifts a barbell from the floor to her waist. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A woman in sleek black activewear performs a series of controlled barbell deadlifts in a well-equipped home gym, her movements precise and deliberate. The camera holds a steady, medium-wide shot, capturing her full body as she stands with feet shoulder-width apart, gripping the barbell with a firm overhand hold. With focused determination etched on her face, she bends at the hips and knees, lowering the barbell smoothly from the floor to her waist, maintaining a straight back and engaged core throughout. The gym is bathed in soft, even lighting that highlights the clean, organized space—treadmills, weight racks, and resistance bands visible in the background. The scene remains consistent, with no changes in framing, camera angle, or lighting, emphasizing the continuity and rhythm of her exercise. Her breath is steady, her muscles taut, as she repeats the motion with unwavering form, the metallic clink of the barbell echoing faintly in the quiet room. The atmosphere is one of quiet intensity, underscoring her commitment and strength.
The dog turns its head further to the right while sitting in the grass.
原始整体描述 · MEV global_prompt.short_caption
A happy Pembroke Welsh Corgi with a reddish-brown and white coat sits in a sunny green field, panting with its tongue out in a static medium shot.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A happy Pembroke Welsh Corgi with a reddish-brown and white coat sits in a sunny green field, panting with its tongue out in a static medium shot. In this continuous shot, The dog turns its head further to the right while sitting in the grass. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A happy Pembroke Welsh Corgi with a rich reddish-brown and white coat sits calmly in a sun-drenched green field, its short legs tucked beneath its sturdy body. The warm sunlight bathes the scene, casting soft highlights across the dog’s smooth fur and illuminating the vibrant blades of grass around it. The corgi pants gently, its pink tongue lolling out in a relaxed, joyful expression, conveying a sense of contentment and ease. The camera holds a steady, static medium shot, framing the dog from the mid-chest up, allowing full view of its expressive face and perked ears. As the moment unfolds, the dog slowly turns its head further to the right, its eyes briefly scanning the surroundings with quiet curiosity, while its body remains still. The lush green field stretches out behind it, dotted with wildflowers and bathed in golden afternoon light. The scene remains consistent throughout—same subject, same environment, same lighting, same framing—capturing a peaceful, serene moment in nature, with the corgi as the calm, cheerful centerpiece.
文本核对：Omni 保留写字动作，但额外加入 occasionally pausing to glance up and around，目标指令没有要求这一动作。
Caption 来源与实际生成文本（展开对照）
目标动作标注 · 与 Editing instruction 的动作正文对应
The young woman sitting on a park bench thoughtfully writes in her notebook with a pen.
原始整体描述 · MEV global_prompt.short_caption
A young woman with shoulder-length light brown hair sits on a wooden bench in a sunny park or campus setting, writing in a notebook while dressed in a pink shirt and blue jeans.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A young woman with shoulder-length light brown hair sits on a wooden bench in a sunny park or campus setting, writing in a notebook while dressed in a pink shirt and blue jeans. In this continuous shot, The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A young woman with shoulder-length light brown hair sits calmly on a weathered wooden bench in a sun-drenched park or campus setting, the golden light of midday filtering through the leaves of nearby trees. She wears a soft pink button-up shirt and fitted blue jeans, her posture relaxed yet focused as she leans slightly forward, pen in hand, writing thoughtfully in a small notebook resting on her lap. Her brow is gently furrowed in concentration, occasionally pausing to glance up and around, as if gathering inspiration from the peaceful surroundings. The camera holds steady in a medium shot, capturing her from the waist up, with the background softly blurred to emphasize her presence—dappled sunlight dances across the grass and pavement, and the faint rustle of leaves and distant chatter of people passing by add a gentle soundtrack to the scene. The composition remains unchanged throughout, maintaining consistent framing, lighting, and perspective, creating a serene, contemplative moment frozen in time.
The woman in a black leather jacket and red pants turns and leans back against the hood of the black car.
原始整体描述 · MEV global_prompt.short_caption
A confident woman in red pants and a black leather jacket poses in front of a black Chevrolet Camaro in a moody, red-lit underground parking garage.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A confident woman in red pants and a black leather jacket poses in front of a black Chevrolet Camaro in a moody, red-lit underground parking garage. In this continuous shot, The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A confident woman stands in front of a sleek black Chevrolet Camaro in a moody, dimly lit underground parking garage, bathed in a deep, ambient red glow that casts dramatic shadows across the concrete walls and floor. She wears a form-fitting black leather jacket and vibrant red pants that contrast sharply with the dark surroundings, drawing the eye to her poised stance. The camera holds steady in a continuous, smooth shot, capturing her as she turns with deliberate grace, her movement fluid and intentional. She leans back against the curved hood of the car, her body angled slightly, one hand resting lightly on the glossy surface while the other hangs relaxed at her side. The red lighting pulses subtly, reflecting off the car’s polished paint and the woman’s jacket, enhancing the cinematic atmosphere. The scene remains tightly framed, with the Camaro’s aggressive lines and the woman’s commanding presence forming a powerful visual composition. The lighting, framing, and camera position remain consistent throughout, emphasizing the stillness and intensity of the moment, as if frozen in time.
The woman in an apron closes her eyes as she tastes from a metal ladle.
原始整体描述 · MEV global_prompt.short_caption
A young woman in a patterned apron stands in a modern kitchen, tasting food from a ladle with a satisfied expression.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A young woman in a patterned apron stands in a modern kitchen, tasting food from a ladle with a satisfied expression. In this continuous shot, The woman in an apron closes her eyes as she tastes from a metal ladle. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A young woman stands in a bright, modern kitchen, her figure centered in a steady, continuous shot. She wears a patterned apron over casual clothing, the fabric soft and slightly rumpled from use. The kitchen is sleek and well-lit, with natural light streaming through a large window, casting a warm glow across stainless steel appliances and clean white countertops. In her hand, she holds a metal ladle, its surface gleaming under the light. She lifts it slowly to her lips, her eyes closing gently as she tastes the food. Her expression softens into quiet satisfaction—lips parting slightly, a subtle smile forming as the flavor registers. The camera remains fixed, capturing the moment with unwavering focus, preserving the stillness of the scene. The lighting remains consistent, highlighting the contours of her face and the texture of the apron. The background remains unchanged—neatly arranged utensils, a pot simmering on the stove, a bowl of fresh herbs on the counter—all contributing to the calm, domestic atmosphere. The shot lingers, emphasizing the sensory experience, the quiet joy of a meal well-prepared, and the simple pleasure of tasting one’s own creation.
文本核对：Omni 保留放手机再翻身的顺序，但额外加入 scrolling or typing 作为前置动作，可能占用短视频中的目标动作时长；是否实际影响执行需看视频。
Caption 来源与实际生成文本（展开对照）
目标动作标注 · 与 Editing instruction 的动作正文对应
The man in bed puts his smartphone on the nightstand and turns over to go to sleep.
原始整体描述 · MEV global_prompt.short_caption
A young man with dyed-blond hair lies in bed in a dark room, illuminated by his phone's blue light, wearing a pink tank top and a nose ring.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A young man with dyed-blond hair lies in bed in a dark room, illuminated by his phone's blue light, wearing a pink tank top and a nose ring. In this continuous shot, The man in bed puts his smartphone on the nightstand and turns over to go to sleep. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A young man with dyed-blond hair lies on his back in a dimly lit bedroom, the only source of light emanating from the cool blue glow of his smartphone screen. He wears a soft pink tank top, the fabric slightly rumpled from lying still, and a small nose ring glints faintly in the low light as he shifts his head. The room is dark, shadows pooling in the corners, with only the faint outline of a nightstand and a pillow visible. The camera holds a steady, fixed position, capturing the scene in a continuous, unbroken shot. The man’s fingers move slowly across the phone’s screen, scrolling or typing with quiet focus. After a moment, he lowers the device, places it gently on the nightstand beside him, and turns onto his side, pulling the blanket slightly higher. His breathing slows, and he settles into sleep, the blue light fading as the phone powers down. The scene remains unchanged in framing, lighting, and composition, maintaining a calm, intimate stillness throughout.
The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water.
原始整体描述 · MEV global_prompt.short_caption
A small, white-spotted jellyfish drifts and pulsates gracefully in serene, deep blue water.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A small, white-spotted jellyfish drifts and pulsates gracefully in serene, deep blue water. In this continuous shot, The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A small, white-spotted jellyfish drifts effortlessly through serene, deep blue water, its translucent bell gently pulsating in rhythmic motion. The delicate, rhythmic contractions of its body propel it forward in a slow, graceful dance, each pulse creating a subtle ripple in the surrounding water. As the bell contracts, it reveals a distinct ring-like shape, a luminous outline that glows faintly against the rich, uniform blue backdrop. The white spots on its body shimmer faintly with the play of light filtering down from above, adding a soft, ethereal contrast to the deep oceanic hue. The scene remains calm and undisturbed, with no sudden movements or changes in the environment. The camera holds steady, maintaining a consistent framing that centers the jellyfish in the frame, capturing its elegant motion in a continuous, uninterrupted shot. The lighting remains soft and diffused, enhancing the tranquil, otherworldly atmosphere of the deep sea. The jellyfish continues its slow, mesmerizing journey, its form and movement unchanged, creating a peaceful and hypnotic visual experience.
Editing instruction Edit the action so that the woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking.
The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking.
原始整体描述 · MEV global_prompt.short_caption
A cheerful woman in a red bikini top and straw sun hat takes a playful selfie against a green screen background in a static medium shot.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A cheerful woman in a red bikini top and straw sun hat takes a playful selfie against a green screen background in a static medium shot. In this continuous shot, The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A cheerful woman stands in a static medium shot against a bright green screen background, her figure centered and clearly defined. She wears a vibrant red bikini top and a wide-brimmed straw sun hat that casts a soft shadow over her face, enhancing the playful, summery vibe. Her expression is lively and animated as she holds her smartphone at arm’s length, capturing a selfie with a mischievous grin. She sticks her tongue out slightly and winks one eye, adding a flirtatious, carefree energy to the moment. Her dark hair peeks out from under the hat, gently swaying as she shifts her weight slightly, maintaining a relaxed yet engaging posture. The lighting is even and bright, highlighting her features and the vivid colors of her outfit without casting harsh shadows. The camera remains completely still throughout, keeping the framing consistent and focused on her upper body and face. The green screen behind her is uniform and uncluttered, providing a clean, neutral backdrop that emphasizes her presence. The scene exudes a lighthearted, fun atmosphere, capturing a spontaneous and joyful moment in a single, continuous take.
文本核对：初始整体 caption 写 vintage-style dress，目标动作写 green skirt and black shirt，服装描述不够一致；Omni 保留两种描述。旋转动作仍出现在扩写文本中。
Caption 来源与实际生成文本（展开对照）
目标动作标注 · 与 Editing instruction 的动作正文对应
The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf.
原始整体描述 · MEV global_prompt.short_caption
A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house. In this continuous shot, The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house, her movements fluid and deliberate under the warm, golden light of late afternoon. She wears a flowing green skirt that swirls with each turn, paired with a fitted black shirt that contrasts elegantly with the soft hues of the surrounding garden. In her hands, she holds a long, sheer scarf that billows like a ribbon in the breeze, catching the sunlight as she spins in a slow, continuous circle. The camera remains steady, capturing her in a wide, stable frame that emphasizes the harmony between her motion and the tranquil setting. The fence, draped in climbing roses and wildflowers, frames the scene with natural beauty, while the weathered wooden house in the background adds a sense of timeless charm. The lighting is soft and diffused, casting gentle shadows and enhancing the nostalgic atmosphere. Her expression is serene, eyes closed at times, as if lost in the rhythm of the moment. The entire scene unfolds in a single, unbroken shot, with no cuts or shifts in perspective, allowing the viewer to fully immerse in the grace and stillness of the moment.
The man lying on the couch rubs his tired eyes and covers his face with both hands.
原始整体描述 · MEV global_prompt.short_caption
A young man with dark skin, dreadlocks, and a gold septum ring lies on a dark couch in a dimly lit room, showing signs of distress.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A young man with dark skin, dreadlocks, and a gold septum ring lies on a dark couch in a dimly lit room, showing signs of distress. In this continuous shot, The man lying on the couch rubs his tired eyes and covers his face with both hands. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A young man with dark skin and long dreadlocks lies sprawled across a dark, plush couch in a dimly lit room, the low ambient light casting deep shadows that pool around him. A gold septum ring glints faintly as he shifts slightly, his body tense with quiet unease. The room is still and intimate, the atmosphere heavy with a sense of emotional weight. He lifts his hands slowly, rubbing his tired eyes with the heels of his palms, his fingers pressing into the sockets as if trying to push back exhaustion or pain. Then, with a slow, deliberate motion, he covers his face entirely, pressing both hands against his forehead and cheeks, his breath shallow and uneven. The camera holds steady, capturing the moment in a continuous, unbroken shot—no movement, no cut—allowing the viewer to sit with the man’s distress. The framing remains tight and consistent, focusing on his upper body and face, emphasizing the subtle tremors in his hands and the way his shoulders rise and fall with each labored breath. The lighting remains unchanged, soft and moody, highlighting the contours of his face and the texture of his hair, while the surrounding darkness absorbs all distractions, drawing attention solely to his internal struggle. The scene is quiet, almost still, yet charged with unspoken emotion.
The woman turns her head to the left and places her water bottle onto the treadmill.
原始整体描述 · MEV global_prompt.short_caption
A young woman with long reddish-blonde hair, wearing a mint green athletic outfit, stands on a treadmill in a modern gym with blue and white striped pillars, holding a water bottle.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A young woman with long reddish-blonde hair, wearing a mint green athletic outfit, stands on a treadmill in a modern gym with blue and white striped pillars, holding a water bottle. In this continuous shot, The woman turns her head to the left and places her water bottle onto the treadmill. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A young woman with long, sun-kissed reddish-blonde hair stands centered on a treadmill in a sleek, modern gym, her posture relaxed yet alert. She wears a form-fitting mint green athletic outfit—a moisture-wicking tank top and matching leggings—that contrasts softly against the cool-toned environment. The gym is illuminated by bright, even lighting that highlights the clean lines of the space, with blue and white striped pillars rising in the background, creating a rhythmic visual pattern that adds depth to the scene. The camera holds a steady, medium-wide shot, capturing her from the waist up, with the treadmill’s control panel and running belt clearly visible. In a smooth, continuous motion, she turns her head slightly to the left, her gaze briefly shifting toward the side of the treadmill, then lowers her right hand to place a clear water bottle with a blue cap onto the treadmill’s console. The bottle rests securely, and she remains in place, her expression calm and focused, as the scene continues without interruption. The lighting remains consistent, the framing stable, and the background elements—pillars, mirrors, and distant gym equipment—remain unchanged, emphasizing the quiet, deliberate nature of the moment.
The woman in a taekwondo uniform stands in profile and claps her hands together in front of her.
原始整体描述 · MEV global_prompt.short_caption
A woman in a white taekwondo uniform with a black collar and Adidas logo stands in profile against a black background under dramatic side lighting, captured in a static medium shot.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A woman in a white taekwondo uniform with a black collar and Adidas logo stands in profile against a black background under dramatic side lighting, captured in a static medium shot. In this continuous shot, The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A woman in a crisp white taekwondo uniform stands in profile against a deep black background, her form sharply defined by dramatic side lighting that casts a strong, sculpted shadow along her silhouette. The uniform features a black collar and a subtle Adidas logo on the chest, adding a touch of brand detail to the athletic attire. She is captured in a static medium shot, framed with precision to highlight her poised stance and the clean lines of her posture. The lighting emphasizes the contours of her body, creating a high-contrast, cinematic effect that enhances the intensity of the moment. In a smooth, continuous motion, she brings her hands together in front of her chest, clapping them once with a deliberate, controlled gesture. Her movements are precise and measured, reflecting discipline and focus. The scene remains unchanged throughout—no shifts in camera angle, lighting, or background—maintaining a consistent, composed atmosphere that centers entirely on her presence and action. The stillness of the frame contrasts with the subtle energy of her motion, creating a powerful visual statement of control and readiness.
d908552fb691 · seed 2026081903 · 81 frames / 25 fps · T2V 运行记录 · Omni 文本规划结果 · Real target 来源
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-extra6/index.html b/projects/video-action-editing/omni-extra6/index.html
new file mode 100644
index 0000000..346d31f
--- /dev/null
+++ b/projects/video-action-editing/omni-extra6/index.html
@@ -0,0 +1,23 @@
+
+OmniVideo2 · 追加六例对比
文本核对：Omni 扩写把 lifts from the floor to her waist 写成 lowering ... from the floor to her waist，动作方向语义矛盾。初始整体 caption 还描述重复 deadlifts。
Caption 来源与实际生成文本（展开对照）
目标动作标注 · 与 Editing instruction 的动作正文对应
The woman in a home gym lifts a barbell from the floor to her waist.
原始整体描述 · MEV global_prompt.short_caption
A woman in black activewear performs barbell deadlifts with focused determination in a well-equipped home gym.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A woman in black activewear performs barbell deadlifts with focused determination in a well-equipped home gym. In this continuous shot, The woman in a home gym lifts a barbell from the floor to her waist. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A woman in sleek black activewear performs a series of controlled barbell deadlifts in a well-equipped home gym, her movements precise and deliberate. The camera holds a steady, medium-wide shot, capturing her full body as she stands with feet shoulder-width apart, gripping the barbell with a firm overhand hold. With focused determination etched on her face, she bends at the hips and knees, lowering the barbell smoothly from the floor to her waist, maintaining a straight back and engaged core throughout. The gym is bathed in soft, even lighting that highlights the clean, organized space—treadmills, weight racks, and resistance bands visible in the background. The scene remains consistent, with no changes in framing, camera angle, or lighting, emphasizing the continuity and rhythm of her exercise. Her breath is steady, her muscles taut, as she repeats the motion with unwavering form, the metallic clink of the barbell echoing faintly in the quiet room. The atmosphere is one of quiet intensity, underscoring her commitment and strength.
The dog turns its head further to the right while sitting in the grass.
原始整体描述 · MEV global_prompt.short_caption
A happy Pembroke Welsh Corgi with a reddish-brown and white coat sits in a sunny green field, panting with its tongue out in a static medium shot.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A happy Pembroke Welsh Corgi with a reddish-brown and white coat sits in a sunny green field, panting with its tongue out in a static medium shot. In this continuous shot, The dog turns its head further to the right while sitting in the grass. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A happy Pembroke Welsh Corgi with a rich reddish-brown and white coat sits calmly in a sun-drenched green field, its short legs tucked beneath its sturdy body. The warm sunlight bathes the scene, casting soft highlights across the dog’s smooth fur and illuminating the vibrant blades of grass around it. The corgi pants gently, its pink tongue lolling out in a relaxed, joyful expression, conveying a sense of contentment and ease. The camera holds a steady, static medium shot, framing the dog from the mid-chest up, allowing full view of its expressive face and perked ears. As the moment unfolds, the dog slowly turns its head further to the right, its eyes briefly scanning the surroundings with quiet curiosity, while its body remains still. The lush green field stretches out behind it, dotted with wildflowers and bathed in golden afternoon light. The scene remains consistent throughout—same subject, same environment, same lighting, same framing—capturing a peaceful, serene moment in nature, with the corgi as the calm, cheerful centerpiece.
文本核对：Omni 保留写字动作，但额外加入 occasionally pausing to glance up and around，目标指令没有要求这一动作。
Caption 来源与实际生成文本（展开对照）
目标动作标注 · 与 Editing instruction 的动作正文对应
The young woman sitting on a park bench thoughtfully writes in her notebook with a pen.
原始整体描述 · MEV global_prompt.short_caption
A young woman with shoulder-length light brown hair sits on a wooden bench in a sunny park or campus setting, writing in a notebook while dressed in a pink shirt and blue jeans.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A young woman with shoulder-length light brown hair sits on a wooden bench in a sunny park or campus setting, writing in a notebook while dressed in a pink shirt and blue jeans. In this continuous shot, The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A young woman with shoulder-length light brown hair sits calmly on a weathered wooden bench in a sun-drenched park or campus setting, the golden light of midday filtering through the leaves of nearby trees. She wears a soft pink button-up shirt and fitted blue jeans, her posture relaxed yet focused as she leans slightly forward, pen in hand, writing thoughtfully in a small notebook resting on her lap. Her brow is gently furrowed in concentration, occasionally pausing to glance up and around, as if gathering inspiration from the peaceful surroundings. The camera holds steady in a medium shot, capturing her from the waist up, with the background softly blurred to emphasize her presence—dappled sunlight dances across the grass and pavement, and the faint rustle of leaves and distant chatter of people passing by add a gentle soundtrack to the scene. The composition remains unchanged throughout, maintaining consistent framing, lighting, and perspective, creating a serene, contemplative moment frozen in time.
The woman in a black leather jacket and red pants turns and leans back against the hood of the black car.
原始整体描述 · MEV global_prompt.short_caption
A confident woman in red pants and a black leather jacket poses in front of a black Chevrolet Camaro in a moody, red-lit underground parking garage.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A confident woman in red pants and a black leather jacket poses in front of a black Chevrolet Camaro in a moody, red-lit underground parking garage. In this continuous shot, The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A confident woman stands in front of a sleek black Chevrolet Camaro in a moody, dimly lit underground parking garage, bathed in a deep, ambient red glow that casts dramatic shadows across the concrete walls and floor. She wears a form-fitting black leather jacket and vibrant red pants that contrast sharply with the dark surroundings, drawing the eye to her poised stance. The camera holds steady in a continuous, smooth shot, capturing her as she turns with deliberate grace, her movement fluid and intentional. She leans back against the curved hood of the car, her body angled slightly, one hand resting lightly on the glossy surface while the other hangs relaxed at her side. The red lighting pulses subtly, reflecting off the car’s polished paint and the woman’s jacket, enhancing the cinematic atmosphere. The scene remains tightly framed, with the Camaro’s aggressive lines and the woman’s commanding presence forming a powerful visual composition. The lighting, framing, and camera position remain consistent throughout, emphasizing the stillness and intensity of the moment, as if frozen in time.
The woman in an apron closes her eyes as she tastes from a metal ladle.
原始整体描述 · MEV global_prompt.short_caption
A young woman in a patterned apron stands in a modern kitchen, tasting food from a ladle with a satisfied expression.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A young woman in a patterned apron stands in a modern kitchen, tasting food from a ladle with a satisfied expression. In this continuous shot, The woman in an apron closes her eyes as she tastes from a metal ladle. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A young woman stands in a bright, modern kitchen, her figure centered in a steady, continuous shot. She wears a patterned apron over casual clothing, the fabric soft and slightly rumpled from use. The kitchen is sleek and well-lit, with natural light streaming through a large window, casting a warm glow across stainless steel appliances and clean white countertops. In her hand, she holds a metal ladle, its surface gleaming under the light. She lifts it slowly to her lips, her eyes closing gently as she tastes the food. Her expression softens into quiet satisfaction—lips parting slightly, a subtle smile forming as the flavor registers. The camera remains fixed, capturing the moment with unwavering focus, preserving the stillness of the scene. The lighting remains consistent, highlighting the contours of her face and the texture of the apron. The background remains unchanged—neatly arranged utensils, a pot simmering on the stove, a bowl of fresh herbs on the counter—all contributing to the calm, domestic atmosphere. The shot lingers, emphasizing the sensory experience, the quiet joy of a meal well-prepared, and the simple pleasure of tasting one’s own creation.
文本核对：Omni 保留放手机再翻身的顺序，但额外加入 scrolling or typing 作为前置动作，可能占用短视频中的目标动作时长；是否实际影响执行需看视频。
Caption 来源与实际生成文本（展开对照）
目标动作标注 · 与 Editing instruction 的动作正文对应
The man in bed puts his smartphone on the nightstand and turns over to go to sleep.
原始整体描述 · MEV global_prompt.short_caption
A young man with dyed-blond hair lies in bed in a dark room, illuminated by his phone's blue light, wearing a pink tank top and a nose ring.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A young man with dyed-blond hair lies in bed in a dark room, illuminated by his phone's blue light, wearing a pink tank top and a nose ring. In this continuous shot, The man in bed puts his smartphone on the nightstand and turns over to go to sleep. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A young man with dyed-blond hair lies on his back in a dimly lit bedroom, the only source of light emanating from the cool blue glow of his smartphone screen. He wears a soft pink tank top, the fabric slightly rumpled from lying still, and a small nose ring glints faintly in the low light as he shifts his head. The room is dark, shadows pooling in the corners, with only the faint outline of a nightstand and a pillow visible. The camera holds a steady, fixed position, capturing the scene in a continuous, unbroken shot. The man’s fingers move slowly across the phone’s screen, scrolling or typing with quiet focus. After a moment, he lowers the device, places it gently on the nightstand beside him, and turns onto his side, pulling the blanket slightly higher. His breathing slows, and he settles into sleep, the blue light fading as the phone powers down. The scene remains unchanged in framing, lighting, and composition, maintaining a calm, intimate stillness throughout.
The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water.
原始整体描述 · MEV global_prompt.short_caption
A small, white-spotted jellyfish drifts and pulsates gracefully in serene, deep blue water.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A small, white-spotted jellyfish drifts and pulsates gracefully in serene, deep blue water. In this continuous shot, The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A small, white-spotted jellyfish drifts effortlessly through serene, deep blue water, its translucent bell gently pulsating in rhythmic motion. The delicate, rhythmic contractions of its body propel it forward in a slow, graceful dance, each pulse creating a subtle ripple in the surrounding water. As the bell contracts, it reveals a distinct ring-like shape, a luminous outline that glows faintly against the rich, uniform blue backdrop. The white spots on its body shimmer faintly with the play of light filtering down from above, adding a soft, ethereal contrast to the deep oceanic hue. The scene remains calm and undisturbed, with no sudden movements or changes in the environment. The camera holds steady, maintaining a consistent framing that centers the jellyfish in the frame, capturing its elegant motion in a continuous, uninterrupted shot. The lighting remains soft and diffused, enhancing the tranquil, otherworldly atmosphere of the deep sea. The jellyfish continues its slow, mesmerizing journey, its form and movement unchanged, creating a peaceful and hypnotic visual experience.
Editing instruction Edit the action so that the woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking.
The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking.
原始整体描述 · MEV global_prompt.short_caption
A cheerful woman in a red bikini top and straw sun hat takes a playful selfie against a green screen background in a static medium shot.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A cheerful woman in a red bikini top and straw sun hat takes a playful selfie against a green screen background in a static medium shot. In this continuous shot, The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A cheerful woman stands in a static medium shot against a bright green screen background, her figure centered and clearly defined. She wears a vibrant red bikini top and a wide-brimmed straw sun hat that casts a soft shadow over her face, enhancing the playful, summery vibe. Her expression is lively and animated as she holds her smartphone at arm’s length, capturing a selfie with a mischievous grin. She sticks her tongue out slightly and winks one eye, adding a flirtatious, carefree energy to the moment. Her dark hair peeks out from under the hat, gently swaying as she shifts her weight slightly, maintaining a relaxed yet engaging posture. The lighting is even and bright, highlighting her features and the vivid colors of her outfit without casting harsh shadows. The camera remains completely still throughout, keeping the framing consistent and focused on her upper body and face. The green screen behind her is uniform and uncluttered, providing a clean, neutral backdrop that emphasizes her presence. The scene exudes a lighthearted, fun atmosphere, capturing a spontaneous and joyful moment in a single, continuous take.
文本核对：初始整体 caption 写 vintage-style dress，目标动作写 green skirt and black shirt，服装描述不够一致；Omni 保留两种描述。旋转动作仍出现在扩写文本中。
Caption 来源与实际生成文本（展开对照）
目标动作标注 · 与 Editing instruction 的动作正文对应
The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf.
原始整体描述 · MEV global_prompt.short_caption
A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house. In this continuous shot, The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house, her movements fluid and deliberate under the warm, golden light of late afternoon. She wears a flowing green skirt that swirls with each turn, paired with a fitted black shirt that contrasts elegantly with the soft hues of the surrounding garden. In her hands, she holds a long, sheer scarf that billows like a ribbon in the breeze, catching the sunlight as she spins in a slow, continuous circle. The camera remains steady, capturing her in a wide, stable frame that emphasizes the harmony between her motion and the tranquil setting. The fence, draped in climbing roses and wildflowers, frames the scene with natural beauty, while the weathered wooden house in the background adds a sense of timeless charm. The lighting is soft and diffused, casting gentle shadows and enhancing the nostalgic atmosphere. Her expression is serene, eyes closed at times, as if lost in the rhythm of the moment. The entire scene unfolds in a single, unbroken shot, with no cuts or shifts in perspective, allowing the viewer to fully immerse in the grace and stillness of the moment.
The man lying on the couch rubs his tired eyes and covers his face with both hands.
原始整体描述 · MEV global_prompt.short_caption
A young man with dark skin, dreadlocks, and a gold septum ring lies on a dark couch in a dimly lit room, showing signs of distress.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A young man with dark skin, dreadlocks, and a gold septum ring lies on a dark couch in a dimly lit room, showing signs of distress. In this continuous shot, The man lying on the couch rubs his tired eyes and covers his face with both hands. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A young man with dark skin and long dreadlocks lies sprawled across a dark, plush couch in a dimly lit room, the low ambient light casting deep shadows that pool around him. A gold septum ring glints faintly as he shifts slightly, his body tense with quiet unease. The room is still and intimate, the atmosphere heavy with a sense of emotional weight. He lifts his hands slowly, rubbing his tired eyes with the heels of his palms, his fingers pressing into the sockets as if trying to push back exhaustion or pain. Then, with a slow, deliberate motion, he covers his face entirely, pressing both hands against his forehead and cheeks, his breath shallow and uneven. The camera holds steady, capturing the moment in a continuous, unbroken shot—no movement, no cut—allowing the viewer to sit with the man’s distress. The framing remains tight and consistent, focusing on his upper body and face, emphasizing the subtle tremors in his hands and the way his shoulders rise and fall with each labored breath. The lighting remains unchanged, soft and moody, highlighting the contours of his face and the texture of his hair, while the surrounding darkness absorbs all distractions, drawing attention solely to his internal struggle. The scene is quiet, almost still, yet charged with unspoken emotion.
The woman turns her head to the left and places her water bottle onto the treadmill.
原始整体描述 · MEV global_prompt.short_caption
A young woman with long reddish-blonde hair, wearing a mint green athletic outfit, stands on a treadmill in a modern gym with blue and white striped pillars, holding a water bottle.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A young woman with long reddish-blonde hair, wearing a mint green athletic outfit, stands on a treadmill in a modern gym with blue and white striped pillars, holding a water bottle. In this continuous shot, The woman turns her head to the left and places her water bottle onto the treadmill. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A young woman with long, sun-kissed reddish-blonde hair stands centered on a treadmill in a sleek, modern gym, her posture relaxed yet alert. She wears a form-fitting mint green athletic outfit—a moisture-wicking tank top and matching leggings—that contrasts softly against the cool-toned environment. The gym is illuminated by bright, even lighting that highlights the clean lines of the space, with blue and white striped pillars rising in the background, creating a rhythmic visual pattern that adds depth to the scene. The camera holds a steady, medium-wide shot, capturing her from the waist up, with the treadmill’s control panel and running belt clearly visible. In a smooth, continuous motion, she turns her head slightly to the left, her gaze briefly shifting toward the side of the treadmill, then lowers her right hand to place a clear water bottle with a blue cap onto the treadmill’s console. The bottle rests securely, and she remains in place, her expression calm and focused, as the scene continues without interruption. The lighting remains consistent, the framing stable, and the background elements—pillars, mirrors, and distant gym equipment—remain unchanged, emphasizing the quiet, deliberate nature of the moment.
The woman in a taekwondo uniform stands in profile and claps her hands together in front of her.
原始整体描述 · MEV global_prompt.short_caption
A woman in a white taekwondo uniform with a black collar and Adidas logo stands in profile against a black background under dramatic side lighting, captured in a static medium shot.
Bernini / Omni 共同初始 caption · 整体描述 + 目标动作 + 保持约束
A woman in a white taekwondo uniform with a black collar and Adidas logo stands in profile against a black background under dramatic side lighting, captured in a static medium shot. In this continuous shot, The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. The same subject identity, scene, lighting, framing, and camera remain stable.
Omni Qwen 扩写后 caption · 实际送入 T5 的文本
A woman in a crisp white taekwondo uniform stands in profile against a deep black background, her form sharply defined by dramatic side lighting that casts a strong, sculpted shadow along her silhouette. The uniform features a black collar and a subtle Adidas logo on the chest, adding a touch of brand detail to the athletic attire. She is captured in a static medium shot, framed with precision to highlight her poised stance and the clean lines of her posture. The lighting emphasizes the contours of her body, creating a high-contrast, cinematic effect that enhances the intensity of the moment. In a smooth, continuous motion, she brings her hands together in front of her chest, clapping them once with a deliberate, controlled gesture. Her movements are precise and measured, reflecting discipline and focus. The scene remains unchanged throughout—no shifts in camera angle, lighting, or background—maintaining a consistent, composed atmosphere that centers entirely on her presence and action. The stillness of the frame contrasts with the subtle energy of her motion, creating a powerful visual statement of control and readiness.
d908552fb691 · seed 2026081903 · 81 frames / 25 fps · T2V 运行记录 · Omni 文本规划结果 · Real target 来源
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/bernini_receipt.json
new file mode 100644
index 0000000..e007c7f
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"865eda9a9146ff00a53995627229c6241aafdbdbf77716eba1f9a5b47262038d","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"0aac5508e42d41bb6875a63dfc05c610419d38e78c7e13d5bb9b5f1cde028b5b","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"3cee34ecc548972476540e65df8873b1bb47ae8aac4edb8a63042fb6378c1bd9","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"7f3ff14de58e3442beb06a9ef0f9514d44414935072dd967e3c13d014c648ac2","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"aba1eed7fe1a01b0425607c467cfdb4b30e9a0664ff556f7387ceb604724327f","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"3c4699d013d8e1405f81b66da29632d90623deb025eb6c8fee41a0b03f085aa0","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"a0d02f8df855e46c8edb4c8deff6071f188f1fca0a1b2e327612c50990b15c6e","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"736a7da7ba6187f82456db81ff453a92e70d880a44b86eeadb14ea55bb033f38","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"a81ba83a4aefbcfc6f10bc36288bfcfd6691c4d04a53fabf08d2def86557d79c","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"de038db05466227ffa3c40eeb9044e1a6e1820372caf2ee6a2d9adeeee674ba5","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"494305d54d8aa21e7e7f690ada8d2d541712eb0b483b18e9ebf8139b5ad1bf20","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"74df7987cb4ba137be84ed38b097cedf177914623b908b9d682e48ec2584e55b","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"4c7e622e97cc75521e6b359dfefda17f96b7de59df3c3eada7e7675ae368a7f6","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"592ba99d6a33ecd3e0cea9f10fca78dbce3f7224a4300b4ef2d9655f40865b25","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"0b85acb7370cba219f192a623c1f2b6f70d696ac59ac15cdc5a19b42073ca2a3","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"3b24517189012d2cf7fb7805850eed520865d3e7f72767ab22ab2144abe538e6","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"0b85acb7370cba219f192a623c1f2b6f70d696ac59ac15cdc5a19b42073ca2a3","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081910,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"3b24517189012d2cf7fb7805850eed520865d3e7f72767ab22ab2144abe538e6","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"1e77c693a0220e4ea8bfd32fb70439b9168f942737bb1539b36e3cd2d656584a","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"3b24517189012d2cf7fb7805850eed520865d3e7f72767ab22ab2144abe538e6"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"0b85acb7370cba219f192a623c1f2b6f70d696ac59ac15cdc5a19b42073ca2a3","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"3b24517189012d2cf7fb7805850eed520865d3e7f72767ab22ab2144abe538e6","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"0b85acb7370cba219f192a623c1f2b6f70d696ac59ac15cdc5a19b42073ca2a3","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081910,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"3b24517189012d2cf7fb7805850eed520865d3e7f72767ab22ab2144abe538e6","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"9951fa0c852fd8da6cdd33702853b4f51ba4f19707065125387cc32098f0e0fd","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"3b24517189012d2cf7fb7805850eed520865d3e7f72767ab22ab2144abe538e6"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":287,"action_prompt_utf8_sha256":"2ee4f1b1b4989f42c2a1b41aca7d635615860e252de87560f5bad2848688716f","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/10af6c651dbc/source-exact81.mp4","source_video_sha256":"ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"53b3ecfc7e4193e63ccd1554e832c2ecdaf4f5d3c9011965e69a667b7d6d639e","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/rv2v.mp4","sha256":"77f3f0b9446ee0b73d243bc0f5bc4ef6a43854931c4d1d03ce2f7e2b2a4e5f93","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"477f08e21106f4d2ebb214dee271325fe9dafd9d4727cf0a4c311a96d2321718","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/t2v.mp4","sha256":"b169af23b7eb5f9fd796fb01a5fd01421da15d40803d9a996fef4566722b7890","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb","original_pre_snapshot_sha256":"ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"ea86ef53e403f1cc602594cc3623a2fb66fb9b2f65c1461ef2bf62bee82a35fb","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"27bf7fa547c03f8269be8610eb7d937630813038545d9c085e1e2a226558817f","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"7cad9858365bcf3076091294ce24efbfc20feb97804917cf4287ad7d805cd23d","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"858596afad9a48d2306356d9d7d1cd9c27b1fded21b66b4d8e2fdb44c6decacb","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081910,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081910,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"1739f8b9594637c14fc72a8d927b5a9272dba82c435b159abffbd2bd38920acb","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/bernini_t2v.jpg
new file mode 100644
index 0000000..c0c20d3
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/bernini_t2v.mp4
new file mode 100644
index 0000000..47e667e
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/launch.json b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/launch.json
new file mode 100644
index 0000000..9e7dd26
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 0,
+ "case_id": "10af6c651dbc",
+ "name": "Jellyfish contraction",
+ "page": 5,
+ "seed": 2026081910,
+ "prompt": "A small, white-spotted jellyfish drifts and pulsates gracefully in serene, deep blue water. In this continuous shot, The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "2ee4f1b1b4989f42c2a1b41aca7d635615860e252de87560f5bad2848688716f",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/t2v.mp4",
+ "bernini_t2v_sha256": "b169af23b7eb5f9fd796fb01a5fd01421da15d40803d9a996fef4566722b7890",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/10af6c651dbc/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081910",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/10af6c651dbc/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "704514c4b9833292fb5bb1308eeed7454747e161b49564dd10bb3c6fbdb0b638",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:06:32.114519+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081910
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/omni_t2v.jpg
new file mode 100644
index 0000000..5578e58
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/omni_t2v.mp4
new file mode 100644
index 0000000..349eced
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/real_target.jpg
new file mode 100644
index 0000000..dc8143c
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/real_target.mp4
new file mode 100644
index 0000000..0581e04
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/real_target_receipt.json
new file mode 100644
index 0000000..61024b4
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "10af6c651dbc",
+ "collected_utc": "2026-09-23T17:34:58.554119+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/853e5739-1635-4de1-8da0-3288460ccb98-00.00.01.720-00.00.07.000-seg2.mp4",
+ "sha256": "b0a83cb3001fceea40d4251c7f747a67f600ff361f34939db767d0ee440aa790"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "24/1",
+ "duration": "5.291667",
+ "nb_read_frames": "127"
+ },
+ "display": {
+ "path": "results/10af6c651dbc/real_target.mp4",
+ "sha256": "c848feb636d4f40415224a26f79cc88659a34d63506b1ea2d95fecb6a7a01542",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/10af6c651dbc.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*0.612283425997894,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/10af6c651dbc/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/receipt.json b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/receipt.json
new file mode 100644
index 0000000..4ffede2
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 0,
+ "case_id": "10af6c651dbc",
+ "name": "Jellyfish contraction",
+ "page": 5,
+ "seed": 2026081910,
+ "prompt": "A small, white-spotted jellyfish drifts and pulsates gracefully in serene, deep blue water. In this continuous shot, The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "2ee4f1b1b4989f42c2a1b41aca7d635615860e252de87560f5bad2848688716f",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/t2v.mp4",
+ "bernini_t2v_sha256": "b169af23b7eb5f9fd796fb01a5fd01421da15d40803d9a996fef4566722b7890",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/10af6c651dbc/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/10af6c651dbc/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081910",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/10af6c651dbc/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "704514c4b9833292fb5bb1308eeed7454747e161b49564dd10bb3c6fbdb0b638",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:06:32.114519+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081910
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A small, white-spotted jellyfish drifts and pulsates gracefully in serene, deep blue water. In this continuous shot, The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A small, white-spotted jellyfish drifts effortlessly through serene, deep blue water, its translucent bell gently pulsating in rhythmic motion. The delicate, rhythmic contractions of its body propel it forward in a slow, graceful dance, each pulse creating a subtle ripple in the surrounding water. As the bell contracts, it reveals a distinct ring-like shape, a luminous outline that glows faintly against the rich, uniform blue backdrop. The white spots on its body shimmer faintly with the play of light filtering down from above, adding a soft, ethereal contrast to the deep oceanic hue. The scene remains calm and undisturbed, with no sudden movements or changes in the environment. The camera holds steady, maintaining a consistent framing that centers the jellyfish in the frame, capturing its elegant motion in a continuous, uninterrupted shot. The lighting remains soft and diffused, enhancing the tranquil, otherworldly atmosphere of the deep sea. The jellyfish continues its slow, mesmerizing journey, its form and movement unchanged, creating a peaceful and hypnotic visual experience.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 67,
+ 2048
+ ],
+ "t5_context_shape": [
+ 253,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 334
+ ],
+ "rank_zero_log_sha256": "1253314f59ddf9f03be7e95cfc2dd23bac0a7d266cbca65d56c57593ab4e1d71"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/10af6c651dbc/omni_t2v.mp4",
+ "sha256": "2380b955b966ff8643bf20f80996659d6c9b2f986743da29a3a2136941ed4437",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 779.17,
+ "completed_utc": "2026-09-23T08:18:48.817827+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/t2v_conditioning.json
new file mode 100644
index 0000000..9c461d3
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/10af6c651dbc/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A small, white-spotted jellyfish drifts and pulsates gracefully in serene, deep blue water. In this continuous shot, The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A small, white-spotted jellyfish drifts effortlessly through serene, deep blue water, its translucent bell gently pulsating in rhythmic motion. The delicate, rhythmic contractions of its body propel it forward in a slow, graceful dance, each pulse creating a subtle ripple in the surrounding water. As the bell contracts, it reveals a distinct ring-like shape, a luminous outline that glows faintly against the rich, uniform blue backdrop. The white spots on its body shimmer faintly with the play of light filtering down from above, adding a soft, ethereal contrast to the deep oceanic hue. The scene remains calm and undisturbed, with no sudden movements or changes in the environment. The camera holds steady, maintaining a consistent framing that centers the jellyfish in the frame, capturing its elegant motion in a continuous, uninterrupted shot. The lighting remains soft and diffused, enhancing the tranquil, otherworldly atmosphere of the deep sea. The jellyfish continues its slow, mesmerizing journey, its form and movement unchanged, creating a peaceful and hypnotic visual experience.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 67,
+ 2048
+ ],
+ "t5_context_shape": [
+ 253,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/bernini_receipt.json
new file mode 100644
index 0000000..0c9931c
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"0e94b25d3a005a1c1fba466437891401b4186fcfa012a82735fd9a30c43cce3c","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"71d7da6d24dd88019b8e2dbfc0274f025aaaaeec606ca658a3f8c31a4f7b101d","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"56d4cf3e697f2c1ca310581171200e937ae8d0ddbc573931d1968e70b585358a","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"0f7bf58fe1fae3e602d7ebf4826023dac24aa6ee89d498bb3c6fab0302bec51b","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"6cf432b68d741603f50739f6ef6d4060769e84c01eb8a077ab5310d5965df28f","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"4afdec2b3a8fae6fcba904e3df789f4aa6d48e11cb60762b79c80728be62f473","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"1b143a73c2c814b5e973c915edcf2526b98f27487323d1ba2751bb775f4c5a0b","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"03d5e456c3bcdd50a38f1aba843ac046b14db363294a45d9cc739c364382419c","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"c04e4d63c40dc181ae413af983a9fab4371c510452c2e759d1e7b565f77325ff","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"88a22b8c0915e0b803703869cc3c3f984b38a42ba50fa6314f2b40e2297439c6","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"94515eedaa3561dbcb7072999e37f233c23ad72c694536d4da01e71073b92dc3","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"21a5afb6de9cf9fc94c35cbedffd4eb09f9a71beb600986351741fd9eadba4be","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"b55c93057cf7c6f7053fa078a44669105d92863bd979bb1c40ce9a4a184b0087","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"2d2bad67a121496c88916f0836f6f5a64c5b349dfd8c2995cfb78e259bff09ee","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"6ea2ae31bea2c05d35ed0dfccb0eb091986304939bdd6e886f411b0b58f7ed5b","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"b7c1abda804811dbfd17180c4cd067d953f3fae3b6aae225c26a6c7e50cc4ed6","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"6ea2ae31bea2c05d35ed0dfccb0eb091986304939bdd6e886f411b0b58f7ed5b","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081902,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"b7c1abda804811dbfd17180c4cd067d953f3fae3b6aae225c26a6c7e50cc4ed6","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"2aa46249b4ff7cfa61b8b4ceac015c9da19a1fe4df741e38bb38b7b90a797477","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"b7c1abda804811dbfd17180c4cd067d953f3fae3b6aae225c26a6c7e50cc4ed6"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"6ea2ae31bea2c05d35ed0dfccb0eb091986304939bdd6e886f411b0b58f7ed5b","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"b7c1abda804811dbfd17180c4cd067d953f3fae3b6aae225c26a6c7e50cc4ed6","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"6ea2ae31bea2c05d35ed0dfccb0eb091986304939bdd6e886f411b0b58f7ed5b","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081902,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"b7c1abda804811dbfd17180c4cd067d953f3fae3b6aae225c26a6c7e50cc4ed6","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"2a2910cadcea003184cc9bff7e23a887d1f6e7a4119a225ba9e91ba989b12476","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"b7c1abda804811dbfd17180c4cd067d953f3fae3b6aae225c26a6c7e50cc4ed6"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":322,"action_prompt_utf8_sha256":"e5a7d338e43779e2faa2383a663e1b1cd80c5e99400a0119b0d3f5dfe1252623","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/3fadbe0c2961/source-exact81.mp4","source_video_sha256":"99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"d64f729b43bde4a7ab47dfb39c1f77c71b8f7aad6f950bb59cfbeed87e5e9604","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/rv2v.mp4","sha256":"12fa9f9adb2f3996849aaa8c3ea958d1c73848bbbc3c28f018377961b16a8d46","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"8c0757b7d9f565685ecf5fe45aeecb61590bad2987f58a3a2c42eedf433a8349","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/t2v.mp4","sha256":"803fc5156774d814787b26a15fd4649c4a6429a7d2ca9b8e89b352461126c924","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364","original_pre_snapshot_sha256":"99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"99170f8a711a3d5a3e930a995561646ac557366cf92f5ded7ac13c6a986f8364","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"4c8b2e4e963aa56d40d038ab70782a8e8a982d842eb80c83a112e3025f01ad34","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"63d1c47fa69475922a71a4de07757c1f9a13eefd4abdc21888e0bfd021ba4098","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"5b166d802da00ec9db2a81368d4da813f83de5cae2786b81f06bf5f477f89462","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081902,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081902,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"26f210745b68f1a3ff1b8444c86d537d622d951a185a350c3d51e7a8f0e0a126","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/bernini_t2v.jpg
new file mode 100644
index 0000000..1672d24
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/bernini_t2v.mp4
new file mode 100644
index 0000000..dd65924
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/launch.json b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/launch.json
new file mode 100644
index 0000000..76a28e2
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 7,
+ "case_id": "3fadbe0c2961",
+ "name": "狗转头 · Dog head turn",
+ "page": null,
+ "seed": 2026081902,
+ "prompt": "A happy Pembroke Welsh Corgi with a reddish-brown and white coat sits in a sunny green field, panting with its tongue out in a static medium shot. In this continuous shot, The dog turns its head further to the right while sitting in the grass. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "e5a7d338e43779e2faa2383a663e1b1cd80c5e99400a0119b0d3f5dfe1252623",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/t2v.mp4",
+ "bernini_t2v_sha256": "803fc5156774d814787b26a15fd4649c4a6429a7d2ca9b8e89b352461126c924",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/3fadbe0c2961/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081902",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/3fadbe0c2961/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "2bfe0e00ce5d62a06e2efe67ea785fb83355a0a7ee60db854c5e5cadc975c60f",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:19:38.914017+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081902
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/omni_t2v.jpg
new file mode 100644
index 0000000..98208b4
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/omni_t2v.mp4
new file mode 100644
index 0000000..23ffd44
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/real_target.jpg
new file mode 100644
index 0000000..9f03e29
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/real_target.mp4
new file mode 100644
index 0000000..7b2049f
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/real_target_receipt.json
new file mode 100644
index 0000000..06c18d8
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "3fadbe0c2961",
+ "collected_utc": "2026-09-23T17:35:15.640575+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/1bb09296-29be-4359-b1d4-03409ed1e30f-00.00.06.110-00.00.09.625-seg2.mp4",
+ "sha256": "1d7d66c1b1ae73688e1d5efa1bd9daac32bc9d376f8c8673e2aa71070f03c7f3"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "25/1",
+ "duration": "3.520000",
+ "nb_read_frames": "88"
+ },
+ "display": {
+ "path": "results/3fadbe0c2961/real_target.mp4",
+ "sha256": "2d97f922dbbaab42d8fb0ff8419410dc7275ddf07d301b1113871e2d651c1e4f",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/3fadbe0c2961.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*0.920454545454546,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/3fadbe0c2961/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/receipt.json b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/receipt.json
new file mode 100644
index 0000000..a63fc62
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 7,
+ "case_id": "3fadbe0c2961",
+ "name": "狗转头 · Dog head turn",
+ "page": null,
+ "seed": 2026081902,
+ "prompt": "A happy Pembroke Welsh Corgi with a reddish-brown and white coat sits in a sunny green field, panting with its tongue out in a static medium shot. In this continuous shot, The dog turns its head further to the right while sitting in the grass. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "e5a7d338e43779e2faa2383a663e1b1cd80c5e99400a0119b0d3f5dfe1252623",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/t2v.mp4",
+ "bernini_t2v_sha256": "803fc5156774d814787b26a15fd4649c4a6429a7d2ca9b8e89b352461126c924",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/3fadbe0c2961/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/3fadbe0c2961/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081902",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/3fadbe0c2961/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "2bfe0e00ce5d62a06e2efe67ea785fb83355a0a7ee60db854c5e5cadc975c60f",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:19:38.914017+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081902
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A happy Pembroke Welsh Corgi with a reddish-brown and white coat sits in a sunny green field, panting with its tongue out in a static medium shot. In this continuous shot, The dog turns its head further to the right while sitting in the grass. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A happy Pembroke Welsh Corgi with a rich reddish-brown and white coat sits calmly in a sun-drenched green field, its short legs tucked beneath its sturdy body. The warm sunlight bathes the scene, casting soft highlights across the dog’s smooth fur and illuminating the vibrant blades of grass around it. The corgi pants gently, its pink tongue lolling out in a relaxed, joyful expression, conveying a sense of contentment and ease. The camera holds a steady, static medium shot, framing the dog from the mid-chest up, allowing full view of its expressive face and perked ears. As the moment unfolds, the dog slowly turns its head further to the right, its eyes briefly scanning the surroundings with quiet curiosity, while its body remains still. The lush green field stretches out behind it, dotted with wildflowers and bathed in golden afternoon light. The scene remains consistent throughout—same subject, same environment, same lighting, same framing—capturing a peaceful, serene moment in nature, with the corgi as the calm, cheerful centerpiece.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 77,
+ 2048
+ ],
+ "t5_context_shape": [
+ 250,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 341
+ ],
+ "rank_zero_log_sha256": "2a6eb521a55ce5cdc05902458c5a2b863cb3532531dd394b6fc5278eac4aa7db"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/3fadbe0c2961/omni_t2v.mp4",
+ "sha256": "6b28c93fda92e5fa29cfe7e1954ee29dd4c6e1d594b657d7a9cf7df195b67e56",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 555.7,
+ "completed_utc": "2026-09-23T08:28:28.823972+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/t2v_conditioning.json
new file mode 100644
index 0000000..d0e86d6
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/3fadbe0c2961/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A happy Pembroke Welsh Corgi with a reddish-brown and white coat sits in a sunny green field, panting with its tongue out in a static medium shot. In this continuous shot, The dog turns its head further to the right while sitting in the grass. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A happy Pembroke Welsh Corgi with a rich reddish-brown and white coat sits calmly in a sun-drenched green field, its short legs tucked beneath its sturdy body. The warm sunlight bathes the scene, casting soft highlights across the dog\u2019s smooth fur and illuminating the vibrant blades of grass around it. The corgi pants gently, its pink tongue lolling out in a relaxed, joyful expression, conveying a sense of contentment and ease. The camera holds a steady, static medium shot, framing the dog from the mid-chest up, allowing full view of its expressive face and perked ears. As the moment unfolds, the dog slowly turns its head further to the right, its eyes briefly scanning the surroundings with quiet curiosity, while its body remains still. The lush green field stretches out behind it, dotted with wildflowers and bathed in golden afternoon light. The scene remains consistent throughout\u2014same subject, same environment, same lighting, same framing\u2014capturing a peaceful, serene moment in nature, with the corgi as the calm, cheerful centerpiece.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 77,
+ 2048
+ ],
+ "t5_context_shape": [
+ 250,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/bernini_receipt.json
new file mode 100644
index 0000000..089c2b5
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"48c0f5010ed6e8b41d140dad22b63aaab5769acd75e26da3d381e0dfa08091d5","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"9ef3aa6493a2dab3675a566a58d9959531e0cf1bad8626f7975749ecda07f21a","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"c5c47a2b972fd914098f76ddd0861a3855edce832f8c25776f542b0c6c2f27b4","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"e3c811659c17cb43c44ac510fe3a9268fc530a5f519386e3f7de6afdbe2f6e44","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"168ad58122626b2a7861516828a7dd487584e4adbf6780821a88dc67f000ce0e","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"bec8e27e605a0471ea304d613ed80ec1de70d3a632e201c29c1ba7daf5386886","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"937d1efd5e00acc1ada41cbff8e3d326cc79aaab9672b02837536c4beb7b2965","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"1bd42ac658d35987dddd4cfbd2b9721fae63ffc3709f9c6dae7956584df73039","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"14c52b34b95baeaf1844052666c85132bb61df9d82146d6de38ebb65d24bfadf","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"0f90a1b7747980f2c82cd1bf605c6019387a2ecb06bae4336f84ff0a9c9b4624","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"b6436ff70eaea0076c99f3c4abd79175b91eed10d328f74a3c04f53d9fdf59ba","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"947f8ca8d2791443c2f23f0e040de2a70cca9ad23f2279af5f90e007ba0b1dcc","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"742bcaf673a42ecba49042671834f363825f3d2f011c2bf3fca90e995fa690f6","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"8bf4d4e4a9b68237f44017bd9c31b985fcfa6e44ccfd79250350e97a89326cd3","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"61bad558235fdfa10344a5355301adbe33a59ed3bd04585bd736f85e5f6cca40","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"28a421cb9991a4dc90b0dec671c069039fa1170741fbf3729124976306436ddf","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"61bad558235fdfa10344a5355301adbe33a59ed3bd04585bd736f85e5f6cca40","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081911,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"28a421cb9991a4dc90b0dec671c069039fa1170741fbf3729124976306436ddf","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"d8a3cc49ff663c598c89eb9b06ccba14567665c08b30989ec7d2ce82575e96f0","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"28a421cb9991a4dc90b0dec671c069039fa1170741fbf3729124976306436ddf"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"61bad558235fdfa10344a5355301adbe33a59ed3bd04585bd736f85e5f6cca40","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"28a421cb9991a4dc90b0dec671c069039fa1170741fbf3729124976306436ddf","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"61bad558235fdfa10344a5355301adbe33a59ed3bd04585bd736f85e5f6cca40","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081911,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"28a421cb9991a4dc90b0dec671c069039fa1170741fbf3729124976306436ddf","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"867c90f55a07c77c58a9d693db5834ced30b880cdeb542be0bc783b2c20a4673","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"28a421cb9991a4dc90b0dec671c069039fa1170741fbf3729124976306436ddf"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":328,"action_prompt_utf8_sha256":"e7cb0b2a1421644429ecb32cb531713267d01c95cc7ec7589a183c2cd226fd21","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/40712e1341dc/source-exact81.mp4","source_video_sha256":"654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"29a9cd1b0eb92704934645b4d4ce2afc2e232817fbb4ba77fefbe9f4cb2e1d97","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/rv2v.mp4","sha256":"5b91b30b5632f6b95cd00198a19f90bb064f16d68d146d39eb680a8ae4169b73","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"397bac948f9549b04fa313b1a3b8fe7f649f1619118cf669accef467e0b1d3fa","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/t2v.mp4","sha256":"99ed719ef9adfa0e959e1cea5af546ba4b61d49d31119a9c28afe301d734bfac","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69","original_pre_snapshot_sha256":"654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"654e997c46a6f78a2ad78359b713767aefe61a6cc732b5be4527720323349f69","source_derived_bucket_hw":[368,656],"source_input_hw":[1080,1920],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"c502d6f1e4920488affe1b0aab2b87b2f27f5a75c8790bb5b3f36a2435dd8420","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"0ffd088bdfbad0beae08af0e57744a4b5c5b4b482e2e3eb4fa0bd6939ba0979c","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"efee3e3f36fedf0dbaf21a08996e77923d30ef158bb6610d30c8e1c1b8ba7b1f","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081911,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081911,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"7294dcc50d25701ed3131e5a7d3eee88e168fcd9904664d6917b057d71dc0314","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/bernini_t2v.jpg
new file mode 100644
index 0000000..576f504
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/bernini_t2v.mp4
new file mode 100644
index 0000000..325879a
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/launch.json b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/launch.json
new file mode 100644
index 0000000..6fd4298
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 11,
+ "case_id": "40712e1341dc",
+ "name": "放下手机后翻身睡觉 · Put phone away and sleep",
+ "page": null,
+ "seed": 2026081911,
+ "prompt": "A young man with dyed-blond hair lies in bed in a dark room, illuminated by his phone's blue light, wearing a pink tank top and a nose ring. In this continuous shot, The man in bed puts his smartphone on the nightstand and turns over to go to sleep. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "e7cb0b2a1421644429ecb32cb531713267d01c95cc7ec7589a183c2cd226fd21",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/t2v.mp4",
+ "bernini_t2v_sha256": "99ed719ef9adfa0e959e1cea5af546ba4b61d49d31119a9c28afe301d734bfac",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/40712e1341dc/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081911",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/40712e1341dc/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "97496b4cd640186777543c0eb1561afdda47c2f7313f1cb51bf1780bb04e00e6",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:28:42.820400+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081911
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/omni_t2v.jpg
new file mode 100644
index 0000000..b2a4b8c
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/omni_t2v.mp4
new file mode 100644
index 0000000..89bb81c
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/real_target.jpg
new file mode 100644
index 0000000..570b836
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/real_target.mp4
new file mode 100644
index 0000000..9c428f3
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/real_target_receipt.json
new file mode 100644
index 0000000..c8e4bb3
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "40712e1341dc",
+ "collected_utc": "2026-09-23T17:35:29.730413+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/88a17729-f1d4-489d-8e73-248bbc088141-00.00.03.347-00.00.06.668-seg2.mp4",
+ "sha256": "2b64f8177ba8494d727bf440238f183a40df32d13d93e8ceb381617063c4fc6a"
+ },
+ "original_probe": {
+ "width": 1920,
+ "height": 1080,
+ "avg_frame_rate": "24000/1001",
+ "duration": "3.336667",
+ "nb_read_frames": "80"
+ },
+ "display": {
+ "path": "results/40712e1341dc/real_target.mp4",
+ "sha256": "4424517beb4e790280cbf782fa1a2d6589b0443d14b9adfb83922d82d919c93e",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/40712e1341dc.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*0.97102887402309,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/40712e1341dc/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/receipt.json b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/receipt.json
new file mode 100644
index 0000000..e759b1b
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 11,
+ "case_id": "40712e1341dc",
+ "name": "放下手机后翻身睡觉 · Put phone away and sleep",
+ "page": null,
+ "seed": 2026081911,
+ "prompt": "A young man with dyed-blond hair lies in bed in a dark room, illuminated by his phone's blue light, wearing a pink tank top and a nose ring. In this continuous shot, The man in bed puts his smartphone on the nightstand and turns over to go to sleep. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "e7cb0b2a1421644429ecb32cb531713267d01c95cc7ec7589a183c2cd226fd21",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/t2v.mp4",
+ "bernini_t2v_sha256": "99ed719ef9adfa0e959e1cea5af546ba4b61d49d31119a9c28afe301d734bfac",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/40712e1341dc/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/40712e1341dc/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081911",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/40712e1341dc/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "97496b4cd640186777543c0eb1561afdda47c2f7313f1cb51bf1780bb04e00e6",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:28:42.820400+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081911
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A young man with dyed-blond hair lies in bed in a dark room, illuminated by his phone's blue light, wearing a pink tank top and a nose ring. In this continuous shot, The man in bed puts his smartphone on the nightstand and turns over to go to sleep. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young man with dyed-blond hair lies on his back in a dimly lit bedroom, the only source of light emanating from the cool blue glow of his smartphone screen. He wears a soft pink tank top, the fabric slightly rumpled from lying still, and a small nose ring glints faintly in the low light as he shifts his head. The room is dark, shadows pooling in the corners, with only the faint outline of a nightstand and a pillow visible. The camera holds a steady, fixed position, capturing the scene in a continuous, unbroken shot. The man’s fingers move slowly across the phone’s screen, scrolling or typing with quiet focus. After a moment, he lowers the device, places it gently on the nightstand beside him, and turns onto his side, pulling the blanket slightly higher. His breathing slows, and he settles into sleep, the blue light fading as the phone powers down. The scene remains unchanged in framing, lighting, and composition, maintaining a calm, intimate stillness throughout.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 80,
+ 2048
+ ],
+ "t5_context_shape": [
+ 239,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 333
+ ],
+ "rank_zero_log_sha256": "762bda731c46287b756b96674e0f3034c5c36417e76e21e33835ff39d69d3781"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/40712e1341dc/omni_t2v.mp4",
+ "sha256": "5ed0844a80af94da135f1290f1307c584e6c5ce29418f8a90f3d940405b2ca11",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 539.42,
+ "completed_utc": "2026-09-23T08:37:24.927593+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/40712e1341dc/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/t2v_conditioning.json
new file mode 100644
index 0000000..2fb6f59
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/40712e1341dc/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A young man with dyed-blond hair lies in bed in a dark room, illuminated by his phone's blue light, wearing a pink tank top and a nose ring. In this continuous shot, The man in bed puts his smartphone on the nightstand and turns over to go to sleep. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young man with dyed-blond hair lies on his back in a dimly lit bedroom, the only source of light emanating from the cool blue glow of his smartphone screen. He wears a soft pink tank top, the fabric slightly rumpled from lying still, and a small nose ring glints faintly in the low light as he shifts his head. The room is dark, shadows pooling in the corners, with only the faint outline of a nightstand and a pillow visible. The camera holds a steady, fixed position, capturing the scene in a continuous, unbroken shot. The man\u2019s fingers move slowly across the phone\u2019s screen, scrolling or typing with quiet focus. After a moment, he lowers the device, places it gently on the nightstand beside him, and turns onto his side, pulling the blanket slightly higher. His breathing slows, and he settles into sleep, the blue light fading as the phone powers down. The scene remains unchanged in framing, lighting, and composition, maintaining a calm, intimate stillness throughout.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 80,
+ 2048
+ ],
+ "t5_context_shape": [
+ 239,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/bernini_receipt.json
new file mode 100644
index 0000000..abf8d9a
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"805525dbeb52b39732e4e78de7ebc47137a6ce356c4f995a3d55603790cd98bc","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"3242f1c9cf2fc266d47b8bc3b91bf48420aa45be7a85a791d7fcc334e4150b2a","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"d83bbf735581805d0152fc360ddda145d74e255521b3f1525df5bb536eb15c78","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"300f82102fc10d09e63e2a17f7c03abc730bd107f7ccb2096126818076aa9af2","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"5f575ff125289c8ee070d4d8b1b8fdef3c9eda1f686592dcfc5b33e7c1054c3b","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"cb49e32bba7aefa6ec60db68e26ae02ebc55e92e7db8d7dcbaaf2b36ba413dcd","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"0258a143a93e3a87b3593858cd7d8bf64aa6101baef2950d3470498593a55560","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"31a1646757cf90a6a221cc036a83ea5b9b2c1c812f282db61fe5da0c81a724e3","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"53f6b977c822c0d9771683cb8ee8ee89c46adf40c807d93cce5e0d125e3bba32","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"94238c183b3a007df492f98fd15a8419bb9cd0b77e414dcd0a5f7aedfc7e53c6","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"6b71e8b1de80e387b54dd229abb4be79f462ea952ce9e3d1b27300f0143b1b97","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"5f73bf575b26aee810a964d6003e5df7008ab2a54435b8e7bb3eaa993903d2a6","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"889b6c48bc39d9658ba758ac4005a288b524064ac800c806546a4e321de5a827","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"a72213df9b5e37ef3c5b85c158a336f6b88dcc5bf9ac8538bf3c99e58881db2c","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"865571d2409a5d20d2f07c752c4d327c6d6bd899c045aaddc4bb8a268a561479","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"0f7c7871f09fd65c843693826404825ab964f51b34034ce985507823187f7cf1","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"865571d2409a5d20d2f07c752c4d327c6d6bd899c045aaddc4bb8a268a561479","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081904,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"0f7c7871f09fd65c843693826404825ab964f51b34034ce985507823187f7cf1","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"32a154b98a2d64c3d8f0ec9a02a549df8f971c2e7bd455834022a0c74a9a2802","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"0f7c7871f09fd65c843693826404825ab964f51b34034ce985507823187f7cf1"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"865571d2409a5d20d2f07c752c4d327c6d6bd899c045aaddc4bb8a268a561479","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"0f7c7871f09fd65c843693826404825ab964f51b34034ce985507823187f7cf1","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"865571d2409a5d20d2f07c752c4d327c6d6bd899c045aaddc4bb8a268a561479","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081904,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"0f7c7871f09fd65c843693826404825ab964f51b34034ce985507823187f7cf1","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"3010a54b7bd926a85e390c9a4a63d1e9f3e2cf9b5ada4bf03cda5dd549ee498f","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"0f7c7871f09fd65c843693826404825ab964f51b34034ce985507823187f7cf1"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":369,"action_prompt_utf8_sha256":"ffe3db596a0db5b6c52685fee0f53d82982bae2a354725ca2eb743558af5df28","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7077cfb160cf/source-exact81.mp4","source_video_sha256":"cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"3348030716a1ebf2eef4e4d9fabc874eb3742b73e2ad6be58b330244168579de","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/rv2v.mp4","sha256":"be645bf32a9ccd63b25e0b4adf4519359be7484f260382d0a3ab46fad28cdb59","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"6913c685c7093d692392138cfbf6c7a98e40132d1fa4e93589b914726db7ec38","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/t2v.mp4","sha256":"c1a8aa80d9a4caa32ccbbc6b00740d576c9d74553361e6686e8d25af98a4645c","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9","original_pre_snapshot_sha256":"cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"cef4c183b459c8aad72cdc81f3e6b2766a36fc22156bf382dd36c29cec6d5aa9","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"3e47bdffb9d5120fad538301af793632e7ad755502d5a14f3b181ab25151e368","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"aa325c341540bec4cbb5474eb77837ce0e5dd17b3d0725a4dc07b234d064ef90","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"2e55a36c23d790d4da7510d98ba6f10cf69770f69b343bcfc0f9c7d639fdc837","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081904,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081904,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"3d0bfb2afb4aa030ed006206e701fd2adef979e7605fe93b34464abf0c9fad97","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/bernini_t2v.jpg
new file mode 100644
index 0000000..3c9229e
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/bernini_t2v.mp4
new file mode 100644
index 0000000..c09d894
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/launch.json b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/launch.json
new file mode 100644
index 0000000..5d193f0
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 8,
+ "case_id": "7077cfb160cf",
+ "name": "长椅上写字 · Write in notebook",
+ "page": null,
+ "seed": 2026081904,
+ "prompt": "A young woman with shoulder-length light brown hair sits on a wooden bench in a sunny park or campus setting, writing in a notebook while dressed in a pink shirt and blue jeans. In this continuous shot, The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "ffe3db596a0db5b6c52685fee0f53d82982bae2a354725ca2eb743558af5df28",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/t2v.mp4",
+ "bernini_t2v_sha256": "c1a8aa80d9a4caa32ccbbc6b00740d576c9d74553361e6686e8d25af98a4645c",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7077cfb160cf/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081904",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7077cfb160cf/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "c108ae74dab9403637ddcb2bb8ca7fca54fa787a68b41d56d8db6bb0805f0352",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:28:45.952216+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081904
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/omni_t2v.jpg
new file mode 100644
index 0000000..751689a
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/omni_t2v.mp4
new file mode 100644
index 0000000..f69d467
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/real_target.jpg
new file mode 100644
index 0000000..f85a241
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/real_target.mp4
new file mode 100644
index 0000000..27cffbd
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/real_target_receipt.json
new file mode 100644
index 0000000..d225a47
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "7077cfb160cf",
+ "collected_utc": "2026-09-23T17:35:14.145966+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/47d49b51-fde1-4c04-a895-4a483df1bfb4-00.00.18.600-00.00.20.229-seg4.mp4",
+ "sha256": "c683137c07b86e3b088c4bd4692b94f1604c84b8bf6a5126dfe9b9a66f4222bf"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "24000/1001",
+ "duration": "1.626625",
+ "nb_read_frames": "39"
+ },
+ "display": {
+ "path": "results/7077cfb160cf/real_target.mp4",
+ "sha256": "03a7b92dc3800e8403ce9f7f4ce85b9668a85a82e110e0356e614d4626496bc3",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/7077cfb160cf.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*1.99185429954661,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/7077cfb160cf/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/receipt.json b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/receipt.json
new file mode 100644
index 0000000..e057a62
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 8,
+ "case_id": "7077cfb160cf",
+ "name": "长椅上写字 · Write in notebook",
+ "page": null,
+ "seed": 2026081904,
+ "prompt": "A young woman with shoulder-length light brown hair sits on a wooden bench in a sunny park or campus setting, writing in a notebook while dressed in a pink shirt and blue jeans. In this continuous shot, The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "ffe3db596a0db5b6c52685fee0f53d82982bae2a354725ca2eb743558af5df28",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/t2v.mp4",
+ "bernini_t2v_sha256": "c1a8aa80d9a4caa32ccbbc6b00740d576c9d74553361e6686e8d25af98a4645c",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7077cfb160cf/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7077cfb160cf/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081904",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7077cfb160cf/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "c108ae74dab9403637ddcb2bb8ca7fca54fa787a68b41d56d8db6bb0805f0352",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:28:45.952216+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081904
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A young woman with shoulder-length light brown hair sits on a wooden bench in a sunny park or campus setting, writing in a notebook while dressed in a pink shirt and blue jeans. In this continuous shot, The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young woman with shoulder-length light brown hair sits calmly on a weathered wooden bench in a sun-drenched park or campus setting, the golden light of midday filtering through the leaves of nearby trees. She wears a soft pink button-up shirt and fitted blue jeans, her posture relaxed yet focused as she leans slightly forward, pen in hand, writing thoughtfully in a small notebook resting on her lap. Her brow is gently furrowed in concentration, occasionally pausing to glance up and around, as if gathering inspiration from the peaceful surroundings. The camera holds steady in a medium shot, capturing her from the waist up, with the background softly blurred to emphasize her presence—dappled sunlight dances across the grass and pavement, and the faint rustle of leaves and distant chatter of people passing by add a gentle soundtrack to the scene. The composition remains unchanged throughout, maintaining consistent framing, lighting, and perspective, creating a serene, contemplative moment frozen in time.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 80,
+ 2048
+ ],
+ "t5_context_shape": [
+ 226,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 320
+ ],
+ "rank_zero_log_sha256": "2974e23fa644b0fa38f0a8091e8bf9fd9ef15ef20ffdb210b14d8a3d1aa75546"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/7077cfb160cf/omni_t2v.mp4",
+ "sha256": "061290c54920d2b552b84b511873f10ead41e881d0f091fc11d0323e97e96427",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 540.8,
+ "completed_utc": "2026-09-23T08:37:29.711466+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/t2v_conditioning.json
new file mode 100644
index 0000000..5098843
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7077cfb160cf/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A young woman with shoulder-length light brown hair sits on a wooden bench in a sunny park or campus setting, writing in a notebook while dressed in a pink shirt and blue jeans. In this continuous shot, The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young woman with shoulder-length light brown hair sits calmly on a weathered wooden bench in a sun-drenched park or campus setting, the golden light of midday filtering through the leaves of nearby trees. She wears a soft pink button-up shirt and fitted blue jeans, her posture relaxed yet focused as she leans slightly forward, pen in hand, writing thoughtfully in a small notebook resting on her lap. Her brow is gently furrowed in concentration, occasionally pausing to glance up and around, as if gathering inspiration from the peaceful surroundings. The camera holds steady in a medium shot, capturing her from the waist up, with the background softly blurred to emphasize her presence\u2014dappled sunlight dances across the grass and pavement, and the faint rustle of leaves and distant chatter of people passing by add a gentle soundtrack to the scene. The composition remains unchanged throughout, maintaining consistent framing, lighting, and perspective, creating a serene, contemplative moment frozen in time.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 80,
+ 2048
+ ],
+ "t5_context_shape": [
+ 226,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/7e488525099e/bernini_receipt.json
new file mode 100644
index 0000000..8f86a2d
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7e488525099e/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"4bb2884085d2efe9bb00f66efadbab32684d79923b32d895e1b73c366a49879b","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"11240a7dc7ab42d595f6f8d2f7b77655426d97744ff4282d776171e1b1e09e8f","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"49af87c18c57a0fe0485c6f1099392d40877bafc82f7df59873313c86341821b","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"4225b445906bad196ebc6211815cf08ba6dfec87d0371b9cc1cf0bda6eaab54a","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"f0f239f7dd0c9de37b9d6946299e406d040691422dc43ddfea09972c13e39579","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"70985e5a80b012470a96ca8d09fd4633814a5b0e6fa2684d1ece78fd678d9038","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"fde7100e71b5abb0fae607bdc626e4ce3c1820b9d05226b300e466e1ff4b442f","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"9e1088117b66b3f7a54ecc0f631a9ad18155d9bce98502e14923a8b74bb83143","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"8debbe0be07c566fc29be4d9cb91d73e63c66f6ced9ce26c0e9fb8e0c3d65a69","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"b4fd87104d2d51303797f63ba0bf14c721f93587933be40526ee440657c14ce5","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"07580c9109678b6dbfdbc5e72916dbc42a62fdb23e4e237a51935aa8b146f616","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"2068aa3ede80a4f0d829b6bf2484cf6682c5324973c4fc23f6e40bc61eb76c16","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"c152613e67559c4812b51cdb049ff7002ca64c315f853864a5e98e8ee5087eaf","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"556b9acb93d700f1b1a5d799a4447cc6ce66e8fdf2ba8944ee981d1552c9c190","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"8080494cfc9ac2566249c4f658b38448d7a1832c5cfbda1fc5461b893895a0e9","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"1f74752f30242d6ce3111ab0c67d57db03fd6d395996fdfd3e9df0b77d358a21","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"8080494cfc9ac2566249c4f658b38448d7a1832c5cfbda1fc5461b893895a0e9","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081906,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"1f74752f30242d6ce3111ab0c67d57db03fd6d395996fdfd3e9df0b77d358a21","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"fd09519ce5c820f7c0f7a9a11bfe5a200ab89eaa4334c5eb1984afc01fd81bfe","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"1f74752f30242d6ce3111ab0c67d57db03fd6d395996fdfd3e9df0b77d358a21"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"8080494cfc9ac2566249c4f658b38448d7a1832c5cfbda1fc5461b893895a0e9","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"1f74752f30242d6ce3111ab0c67d57db03fd6d395996fdfd3e9df0b77d358a21","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"8080494cfc9ac2566249c4f658b38448d7a1832c5cfbda1fc5461b893895a0e9","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081906,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"1f74752f30242d6ce3111ab0c67d57db03fd6d395996fdfd3e9df0b77d358a21","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"18cfca0fdc660fdcd85678294f6e6931f4d68853c12baac8fc1c81d1e8c2ebfd","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"1f74752f30242d6ce3111ab0c67d57db03fd6d395996fdfd3e9df0b77d358a21"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":358,"action_prompt_utf8_sha256":"9dbe71bad14d34165cbbc5236e464eb582094279ee5268bd62cda1d45c15f7a7","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7e488525099e/source-exact81.mp4","source_video_sha256":"1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"19b75c955d886bfe156a837658d89e65cd4b9b84fd80ddd84c856285589087c0","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/rv2v.mp4","sha256":"20154ed35790207b991e7a58e5456b93776d1223c6b666bf37aab5d84396ea08","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"10dc92b7046a38ddf688eded46b7232193947b56887762f3da40d1abebbf2fa2","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/t2v.mp4","sha256":"ef14763c12e45144e6f768a6ed5efae0ccf458086d479815cd12b4832cc17824","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4","original_pre_snapshot_sha256":"1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"1f3cec72bef55b67c15ebfc063252cce867ea128594a225db61688ff961b35d4","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"2767de6b8e73fcf639606514eb89d02fb4b94ffcdd075d5cc8da03ed0772eb85","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"68b01af00e7640dcb5bcfa9690369cfbff48541ab09702eebc53acdd51769f6f","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"b7670ae626c2159e4bef64e612d9fae530782be2c491cac2959a6432d24fca89","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081906,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081906,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"eca1cea19b7a9a1412a53d61d4808256fa6827f49fd779545346b46921e0e771","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/7e488525099e/bernini_t2v.jpg
new file mode 100644
index 0000000..1bfa876
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7e488525099e/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/7e488525099e/bernini_t2v.mp4
new file mode 100644
index 0000000..ea258e0
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7e488525099e/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/launch.json b/projects/video-action-editing/omni-t2v12/results/7e488525099e/launch.json
new file mode 100644
index 0000000..572e7ac
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7e488525099e/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 9,
+ "case_id": "7e488525099e",
+ "name": "转身倚车 · Turn and lean on car",
+ "page": null,
+ "seed": 2026081906,
+ "prompt": "A confident woman in red pants and a black leather jacket poses in front of a black Chevrolet Camaro in a moody, red-lit underground parking garage. In this continuous shot, The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "9dbe71bad14d34165cbbc5236e464eb582094279ee5268bd62cda1d45c15f7a7",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/t2v.mp4",
+ "bernini_t2v_sha256": "ef14763c12e45144e6f768a6ed5efae0ccf458086d479815cd12b4832cc17824",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7e488525099e/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081906",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7e488525099e/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "67b0556dea2138863e9524f2a6a742137d6479fa64a541dbbd88f42de988b8b8",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:06:59.041815+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081906
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/7e488525099e/omni_t2v.jpg
new file mode 100644
index 0000000..7bf0926
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7e488525099e/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/7e488525099e/omni_t2v.mp4
new file mode 100644
index 0000000..6e0468c
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7e488525099e/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/7e488525099e/real_target.jpg
new file mode 100644
index 0000000..c493a04
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7e488525099e/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/7e488525099e/real_target.mp4
new file mode 100644
index 0000000..58ff119
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7e488525099e/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/7e488525099e/real_target_receipt.json
new file mode 100644
index 0000000..241527c
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7e488525099e/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "7e488525099e",
+ "collected_utc": "2026-09-23T17:35:20.553761+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/4fd015eb-922b-4b00-9066-e9d082de9e01-00.00.08.206-00.00.11.256-seg5.mp4",
+ "sha256": "dce487e374dd2fc3c6373d0ceb9e74185a28950791843a13b14d8b2a8a8850cd"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "24000/1001",
+ "duration": "3.128125",
+ "nb_read_frames": "75"
+ },
+ "display": {
+ "path": "results/7e488525099e/real_target.mp4",
+ "sha256": "30c8909ffd3c044c78714102c8d24dc38cf3547c123f0be2deb43390a5732ec3",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/7e488525099e.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*1.03576423576424,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/7e488525099e/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/receipt.json b/projects/video-action-editing/omni-t2v12/results/7e488525099e/receipt.json
new file mode 100644
index 0000000..82300c0
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7e488525099e/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 9,
+ "case_id": "7e488525099e",
+ "name": "转身倚车 · Turn and lean on car",
+ "page": null,
+ "seed": 2026081906,
+ "prompt": "A confident woman in red pants and a black leather jacket poses in front of a black Chevrolet Camaro in a moody, red-lit underground parking garage. In this continuous shot, The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "9dbe71bad14d34165cbbc5236e464eb582094279ee5268bd62cda1d45c15f7a7",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/t2v.mp4",
+ "bernini_t2v_sha256": "ef14763c12e45144e6f768a6ed5efae0ccf458086d479815cd12b4832cc17824",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7e488525099e/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7e488525099e/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081906",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7e488525099e/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "67b0556dea2138863e9524f2a6a742137d6479fa64a541dbbd88f42de988b8b8",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:06:59.041815+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081906
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A confident woman in red pants and a black leather jacket poses in front of a black Chevrolet Camaro in a moody, red-lit underground parking garage. In this continuous shot, The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A confident woman stands in front of a sleek black Chevrolet Camaro in a moody, dimly lit underground parking garage, bathed in a deep, ambient red glow that casts dramatic shadows across the concrete walls and floor. She wears a form-fitting black leather jacket and vibrant red pants that contrast sharply with the dark surroundings, drawing the eye to her poised stance. The camera holds steady in a continuous, smooth shot, capturing her as she turns with deliberate grace, her movement fluid and intentional. She leans back against the curved hood of the car, her body angled slightly, one hand resting lightly on the glossy surface while the other hangs relaxed at her side. The red lighting pulses subtly, reflecting off the car’s polished paint and the woman’s jacket, enhancing the cinematic atmosphere. The scene remains tightly framed, with the Camaro’s aggressive lines and the woman’s commanding presence forming a powerful visual composition. The lighting, framing, and camera position remain consistent throughout, emphasizing the stillness and intensity of the moment, as if frozen in time.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 80,
+ 2048
+ ],
+ "t5_context_shape": [
+ 252,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 346
+ ],
+ "rank_zero_log_sha256": "2cb8628e967b63c4350488a3109cc8bf44c4e79d0bf6ed4828dc3a9486f3b76e"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/7e488525099e/omni_t2v.mp4",
+ "sha256": "bf577e972362491fe79f36079915ac95bd4110ff4a2278184e1c284260ad64e2",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 772.82,
+ "completed_utc": "2026-09-23T08:19:12.510894+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/7e488525099e/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/7e488525099e/t2v_conditioning.json
new file mode 100644
index 0000000..363a6ef
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7e488525099e/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A confident woman in red pants and a black leather jacket poses in front of a black Chevrolet Camaro in a moody, red-lit underground parking garage. In this continuous shot, The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A confident woman stands in front of a sleek black Chevrolet Camaro in a moody, dimly lit underground parking garage, bathed in a deep, ambient red glow that casts dramatic shadows across the concrete walls and floor. She wears a form-fitting black leather jacket and vibrant red pants that contrast sharply with the dark surroundings, drawing the eye to her poised stance. The camera holds steady in a continuous, smooth shot, capturing her as she turns with deliberate grace, her movement fluid and intentional. She leans back against the curved hood of the car, her body angled slightly, one hand resting lightly on the glossy surface while the other hangs relaxed at her side. The red lighting pulses subtly, reflecting off the car\u2019s polished paint and the woman\u2019s jacket, enhancing the cinematic atmosphere. The scene remains tightly framed, with the Camaro\u2019s aggressive lines and the woman\u2019s commanding presence forming a powerful visual composition. The lighting, framing, and camera position remain consistent throughout, emphasizing the stillness and intensity of the moment, as if frozen in time.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 80,
+ 2048
+ ],
+ "t5_context_shape": [
+ 252,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/bernini_receipt.json
new file mode 100644
index 0000000..c87d796
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"f9fb66d7ba26d5b6c907907484ab012f56226413b6589008980e07769bbb72d9","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"8bcc51dc4c2e74329745d5bbb4682931a9e1fc6d0109835d876d3fddb6ab68b7","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"87748a8b55ddaaffbf5218f4f073870cfef1c1640a1f3065e4a4576649f2ef2d","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"c84892f1c13e19eb87fe487d81586dbc534037dce1397517586695c590acf5ee","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"8b35befc7791242197bb98d009552e4b9fd93e824a8a4a8213bce975cb308b50","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"07568fb4d663b6ba3673ec1e193de3821d7d820d6ce1efc00714f841e0eb4a45","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"02b1dd31203f7faf7056896412ab13cc49c1502905eea080dcdef0339f1c78ea","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"989b2ea0df40b524f363fc139658f64999ccc7f337bfbbc20a43312e2d69ca26","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"a89609a93f01479b99e582be450147e79f1f1bb7a743f733f8bb39dbcdf0cd9f","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"e4f73d861568ade971e15f4c2b887157d6a73660c95825bb8dc13ff97664cd20","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"bb2c3cae810ad8747008fc1f29f114d44b35d8ba1e6afea150fad559222261e1","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"622d0aa42aa3d1da17ddf09ad6e2045973df15689519ab42714f674ecf670f79","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"8b0e68030f5e31c73a5aef03712524132ee414e2ef7b72d2498eea16020463de","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"85cb92d76808926625238adaca005b0d57fce9efa62bac4bab9813cada4036b5","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"089225fb799a25a8c375b5fef60937ddc68aecdc9d3b6a48e91d12cab6f53b77","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"bccdd1a64ddd0ab37a65884f8516b175dccce7e4fbfebbf8208748da6825dc4e","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"089225fb799a25a8c375b5fef60937ddc68aecdc9d3b6a48e91d12cab6f53b77","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081900,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"bccdd1a64ddd0ab37a65884f8516b175dccce7e4fbfebbf8208748da6825dc4e","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"a508f70ba841f81c279f0cd4f4af7edd9610bb3b58c79379b1b3d77245b8a0ee","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"bccdd1a64ddd0ab37a65884f8516b175dccce7e4fbfebbf8208748da6825dc4e"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"089225fb799a25a8c375b5fef60937ddc68aecdc9d3b6a48e91d12cab6f53b77","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"bccdd1a64ddd0ab37a65884f8516b175dccce7e4fbfebbf8208748da6825dc4e","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"089225fb799a25a8c375b5fef60937ddc68aecdc9d3b6a48e91d12cab6f53b77","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081900,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"bccdd1a64ddd0ab37a65884f8516b175dccce7e4fbfebbf8208748da6825dc4e","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"b2738639091d9f3e01e0b563099bc817b38d90de3b66bb542ff51e87d5b01db7","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"bccdd1a64ddd0ab37a65884f8516b175dccce7e4fbfebbf8208748da6825dc4e"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":360,"action_prompt_utf8_sha256":"0ff87e5aa29e20f22ba1d2963c2700a3f10b6941147a8e517dd05e339cfb50de","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/7fa338474c75/source-exact81.mp4","source_video_sha256":"4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"14183c91c3d381a24be715255b3c4a9a0f370bcc69c6ed876121f61c18eea7ad","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/rv2v.mp4","sha256":"135d1e590d26a18862564839aa9e228cfde73e2f36162fec1ea410312463efa3","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"672966d65e34461fa1f289c183c94690e098d495348309643d66764adbf0bf19","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/t2v.mp4","sha256":"17b7ccb39f068649eb0ff155d50fdcba822dfd0a66bcf7965086b62651335d34","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840","original_pre_snapshot_sha256":"4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"4278c748fa1fdd54d908c366a781df33d3e4eb1acc1c2b7bd3492659c66a9840","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"390651d2b78ec306d9df7bad9f858f54aad88c2e5c8901777ea14d66e0b557fd","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"b375acf818287eb04d364b4e6659a6a57a3ba76383eff12ea45a12864a3556c0","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"afb24405bd96f305c72df69a42ba8f0b93714080c43bbcc42cb42392c8306099","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081900,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081900,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"e164987cde49343a15ba70910ae003f9287d20d54e99bc618833b536c123e851","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/bernini_t2v.jpg
new file mode 100644
index 0000000..82a6c89
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/bernini_t2v.mp4
new file mode 100644
index 0000000..8704ebe
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/launch.json b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/launch.json
new file mode 100644
index 0000000..f22e920
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 1,
+ "case_id": "7fa338474c75",
+ "name": "Playful selfie",
+ "page": 5,
+ "seed": 2026081900,
+ "prompt": "A cheerful woman in a red bikini top and straw sun hat takes a playful selfie against a green screen background in a static medium shot. In this continuous shot, The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "0ff87e5aa29e20f22ba1d2963c2700a3f10b6941147a8e517dd05e339cfb50de",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/t2v.mp4",
+ "bernini_t2v_sha256": "17b7ccb39f068649eb0ff155d50fdcba822dfd0a66bcf7965086b62651335d34",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7fa338474c75/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081900",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7fa338474c75/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "00d8baf84c06d97b147e7c9a6388d3e7b1ea2fa9c48f787da81820dc29e251d1",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:19:06.632724+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081900
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/omni_t2v.jpg
new file mode 100644
index 0000000..7927e0b
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/omni_t2v.mp4
new file mode 100644
index 0000000..2e2d6e5
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/real_target.jpg
new file mode 100644
index 0000000..a7ad6b9
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/real_target.mp4
new file mode 100644
index 0000000..9f63253
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/real_target_receipt.json
new file mode 100644
index 0000000..cb5b63c
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "7fa338474c75",
+ "collected_utc": "2026-09-23T17:34:57.433253+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/081e2761-c4e1-47ed-8a82-808bec50a59d-00.00.03.040-00.00.08.810-seg2.mp4",
+ "sha256": "d128357efedf0de87647febeadc382f18ee39461fb1e2c3df8de94a5a4578a42"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "60000/1001",
+ "duration": "5.772433",
+ "nb_read_frames": "346"
+ },
+ "display": {
+ "path": "results/7fa338474c75/real_target.mp4",
+ "sha256": "8427b71fda1c01fe061b9943820e618c9f887c03a19f79f9191e4ee879b5a25a",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/7fa338474c75.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*0.561288454972106,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/7fa338474c75/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/receipt.json b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/receipt.json
new file mode 100644
index 0000000..2ac1d17
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 1,
+ "case_id": "7fa338474c75",
+ "name": "Playful selfie",
+ "page": 5,
+ "seed": 2026081900,
+ "prompt": "A cheerful woman in a red bikini top and straw sun hat takes a playful selfie against a green screen background in a static medium shot. In this continuous shot, The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "0ff87e5aa29e20f22ba1d2963c2700a3f10b6941147a8e517dd05e339cfb50de",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/t2v.mp4",
+ "bernini_t2v_sha256": "17b7ccb39f068649eb0ff155d50fdcba822dfd0a66bcf7965086b62651335d34",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/7fa338474c75/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7fa338474c75/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081900",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/7fa338474c75/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "00d8baf84c06d97b147e7c9a6388d3e7b1ea2fa9c48f787da81820dc29e251d1",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:19:06.632724+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081900
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A cheerful woman in a red bikini top and straw sun hat takes a playful selfie against a green screen background in a static medium shot. In this continuous shot, The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A cheerful woman stands in a static medium shot against a bright green screen background, her figure centered and clearly defined. She wears a vibrant red bikini top and a wide-brimmed straw sun hat that casts a soft shadow over her face, enhancing the playful, summery vibe. Her expression is lively and animated as she holds her smartphone at arm’s length, capturing a selfie with a mischievous grin. She sticks her tongue out slightly and winks one eye, adding a flirtatious, carefree energy to the moment. Her dark hair peeks out from under the hat, gently swaying as she shifts her weight slightly, maintaining a relaxed yet engaging posture. The lighting is even and bright, highlighting her features and the vivid colors of her outfit without casting harsh shadows. The camera remains completely still throughout, keeping the framing consistent and focused on her upper body and face. The green screen behind her is uniform and uncluttered, providing a clean, neutral backdrop that emphasizes her presence. The scene exudes a lighthearted, fun atmosphere, capturing a spontaneous and joyful moment in a single, continuous take.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 79,
+ 2048
+ ],
+ "t5_context_shape": [
+ 255,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 348
+ ],
+ "rank_zero_log_sha256": "bfa9dc5a39eb59949e0310fbc226255b473cd5ca45931ddeeb75709e0c0c89bb"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/7fa338474c75/omni_t2v.mp4",
+ "sha256": "b4e27999f2701c65ac62157ae6f49539ce092acc1dba0f652532fe793b037fc3",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 548.75,
+ "completed_utc": "2026-09-23T08:27:57.663754+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/7fa338474c75/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/t2v_conditioning.json
new file mode 100644
index 0000000..caf0daa
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/7fa338474c75/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A cheerful woman in a red bikini top and straw sun hat takes a playful selfie against a green screen background in a static medium shot. In this continuous shot, The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A cheerful woman stands in a static medium shot against a bright green screen background, her figure centered and clearly defined. She wears a vibrant red bikini top and a wide-brimmed straw sun hat that casts a soft shadow over her face, enhancing the playful, summery vibe. Her expression is lively and animated as she holds her smartphone at arm\u2019s length, capturing a selfie with a mischievous grin. She sticks her tongue out slightly and winks one eye, adding a flirtatious, carefree energy to the moment. Her dark hair peeks out from under the hat, gently swaying as she shifts her weight slightly, maintaining a relaxed yet engaging posture. The lighting is even and bright, highlighting her features and the vivid colors of her outfit without casting harsh shadows. The camera remains completely still throughout, keeping the framing consistent and focused on her upper body and face. The green screen behind her is uniform and uncluttered, providing a clean, neutral backdrop that emphasizes her presence. The scene exudes a lighthearted, fun atmosphere, capturing a spontaneous and joyful moment in a single, continuous take.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 79,
+ 2048
+ ],
+ "t5_context_shape": [
+ 255,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/bernini_receipt.json
new file mode 100644
index 0000000..f9dcbf3
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"ce530d73daa6abb5b969f1d3266c691aa435a2c8bd0d4ca9426b5e249f66ef61","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"8f09b8ecfed65bf80b9e34cbc285bfc0cf67767df70140662a0c2840a219c0d7","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"2a53afe1650cfab2d3ac8b51fe33f02235022c4d047947dde7e999f3082191c3","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"95c8c5ec8ad4b3d7233359f321f3315046697aab827a8c951a7ca5a8e652b2f7","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"0cecd0632e41dd74a5c1c077098079bb15da04fb7bf0aa7c73ac5c7732cacb0b","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"8c9f975ba007f1ecdb769a497d9d951978e4757457d760f3c6b17cd6abd1461f","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"c078d8ca086b1ff21cad71b882c029be28e94028d2aa33dc02e84078fc677238","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"8b5b13c7294d4ad3e2be9a775865b71187be53abab4caf09f201710b0a40401d","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"435a03f05de71292fee97d5e6ba138978d92e7737ba7157c4ea326b8e78eeb4f","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"9cc054f3d575a45b39d3c832f67784c2490018697147140dd1c26f75587ff3b4","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"afa5cbf2e8426145cb1086e3c1edd81c012c2eee4462246c2d0317ec511e04e6","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"2f6d3cdd86969ea676d570f08eac7b523a75ea9d3fcbcb4839fb1a33d38c7f2f","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"072a77897e5c0dbfc7897e44a0d9683ae5efaeee64c3445701c68e668e753c6c","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"68905a2905a05bbbe8be8002f5ebc991cc7d11718cf6b0f424164ed8f5aa4091","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"a51983a760145dea7aa4a57824c0130c2be80717592390f360a7819f3f69e229","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"7bdebc631ee914dd17063a860196d41a22461b9486f2082902e591626ea30e10","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"a51983a760145dea7aa4a57824c0130c2be80717592390f360a7819f3f69e229","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081915,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"7bdebc631ee914dd17063a860196d41a22461b9486f2082902e591626ea30e10","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"2295f02c127be7b0749791e4cbc63a2d224283b236635e1fea47789137f3374b","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"7bdebc631ee914dd17063a860196d41a22461b9486f2082902e591626ea30e10"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"a51983a760145dea7aa4a57824c0130c2be80717592390f360a7819f3f69e229","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"7bdebc631ee914dd17063a860196d41a22461b9486f2082902e591626ea30e10","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"a51983a760145dea7aa4a57824c0130c2be80717592390f360a7819f3f69e229","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081915,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"7bdebc631ee914dd17063a860196d41a22461b9486f2082902e591626ea30e10","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"c5f43d5946b20260a431cdcf5de99f7d641e5a66159050aa60c616432821051a","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"7bdebc631ee914dd17063a860196d41a22461b9486f2082902e591626ea30e10"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":313,"action_prompt_utf8_sha256":"f0c58055ee24d894eb9274175c45f1bc41371cd9ff37dbae7d599298eece6bbd","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/81533c9e56ec/source-exact81.mp4","source_video_sha256":"6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"e4f07921d171f20a05139f37be8cb6ac946326e55a06cab5d577bd7a1629211e","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/rv2v.mp4","sha256":"96023f66a9a170c5880e637e2085520de6d8aa2c9889a1dc47a053812644905e","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"971b5d3e7c97d5ad3f00c69c3a637d0b3e9950689a890ab56fcf283be25b218c","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/t2v.mp4","sha256":"7dcf7690d12f8c59a3bef36b2b8a91f681d44351386bbb0f34c9f2a43625be1b","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19","original_pre_snapshot_sha256":"6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"6daa80d3ac407b01f842ce6fa611b3ec6c421851525f0845779143c4577e0b19","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"3806bba0f33f8966ce0ae354187d157fdb3d2bfc2d061326791c9782c36a3bff","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"7ff108fc95afc47faf6136fd7ed84de550b2a1c1db0e8a9f5100ab9be73aa184","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"a431bf93418404503aab9fec5fafe012793198bb0e5e48eb7f234e7640e267c7","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081915,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081915,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"f12cd6af0a68b68a7cf981c67abbbb0af0e4337064c380f67ed1be15d6af48ca","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/bernini_t2v.jpg
new file mode 100644
index 0000000..dc7b150
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/bernini_t2v.mp4
new file mode 100644
index 0000000..60d0e0c
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/launch.json b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/launch.json
new file mode 100644
index 0000000..de3777b
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 2,
+ "case_id": "81533c9e56ec",
+ "name": "Scarf spin",
+ "page": 6,
+ "seed": 2026081915,
+ "prompt": "A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house. In this continuous shot, The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "f0c58055ee24d894eb9274175c45f1bc41371cd9ff37dbae7d599298eece6bbd",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/t2v.mp4",
+ "bernini_t2v_sha256": "7dcf7690d12f8c59a3bef36b2b8a91f681d44351386bbb0f34c9f2a43625be1b",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/81533c9e56ec/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081915",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/81533c9e56ec/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "e0acbe48472ec3cf671de0600e2a962d64aaf95869175e179d42dcce1b9536cf",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:28:15.040774+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081915
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/omni_t2v.jpg
new file mode 100644
index 0000000..6462295
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/omni_t2v.mp4
new file mode 100644
index 0000000..9e59345
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/real_target.jpg
new file mode 100644
index 0000000..9468186
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/real_target.mp4
new file mode 100644
index 0000000..28faef5
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/real_target_receipt.json
new file mode 100644
index 0000000..3a2f4d5
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "81533c9e56ec",
+ "collected_utc": "2026-09-23T17:34:59.431906+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/f810246b-7b74-4eb7-9777-a8f8db92f951-00.00.07.981-00.00.12.766-seg2.mp4",
+ "sha256": "cc70ae83aa709078e90a686426c6ad7956f4d8f722c76811660a4fea4d56373e"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "25/1",
+ "duration": "4.760000",
+ "nb_read_frames": "119"
+ },
+ "display": {
+ "path": "results/81533c9e56ec/real_target.mp4",
+ "sha256": "5edb351b04edd07ccb51fc57043298d62a9abcb959d6f86bd5912ef5d654d6d3",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/81533c9e56ec.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*0.680672268907563,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/81533c9e56ec/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/receipt.json b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/receipt.json
new file mode 100644
index 0000000..3bd80c4
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 2,
+ "case_id": "81533c9e56ec",
+ "name": "Scarf spin",
+ "page": 6,
+ "seed": 2026081915,
+ "prompt": "A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house. In this continuous shot, The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "f0c58055ee24d894eb9274175c45f1bc41371cd9ff37dbae7d599298eece6bbd",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/t2v.mp4",
+ "bernini_t2v_sha256": "7dcf7690d12f8c59a3bef36b2b8a91f681d44351386bbb0f34c9f2a43625be1b",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/81533c9e56ec/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/81533c9e56ec/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081915",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/81533c9e56ec/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "e0acbe48472ec3cf671de0600e2a962d64aaf95869175e179d42dcce1b9536cf",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:28:15.040774+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081915
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house. In this continuous shot, The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house, her movements fluid and deliberate under the warm, golden light of late afternoon. She wears a flowing green skirt that swirls with each turn, paired with a fitted black shirt that contrasts elegantly with the soft hues of the surrounding garden. In her hands, she holds a long, sheer scarf that billows like a ribbon in the breeze, catching the sunlight as she spins in a slow, continuous circle. The camera remains steady, capturing her in a wide, stable frame that emphasizes the harmony between her motion and the tranquil setting. The fence, draped in climbing roses and wildflowers, frames the scene with natural beauty, while the weathered wooden house in the background adds a sense of timeless charm. The lighting is soft and diffused, casting gentle shadows and enhancing the nostalgic atmosphere. Her expression is serene, eyes closed at times, as if lost in the rhythm of the moment. The entire scene unfolds in a single, unbroken shot, with no cuts or shifts in perspective, allowing the viewer to fully immerse in the grace and stillness of the moment.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 67,
+ 2048
+ ],
+ "t5_context_shape": [
+ 274,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 355
+ ],
+ "rank_zero_log_sha256": "b6de7b091cb33a2ef71fff25d838ebe3334d7999fd50c1abde4bbe69a0ff0e80"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/81533c9e56ec/omni_t2v.mp4",
+ "sha256": "11a91b035dac93b7875974b44469c54743c226ea892402e0d742dc90a0e49ca4",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 543.4,
+ "completed_utc": "2026-09-23T08:37:01.166639+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/t2v_conditioning.json
new file mode 100644
index 0000000..41c2440
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/81533c9e56ec/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house. In this continuous shot, The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house, her movements fluid and deliberate under the warm, golden light of late afternoon. She wears a flowing green skirt that swirls with each turn, paired with a fitted black shirt that contrasts elegantly with the soft hues of the surrounding garden. In her hands, she holds a long, sheer scarf that billows like a ribbon in the breeze, catching the sunlight as she spins in a slow, continuous circle. The camera remains steady, capturing her in a wide, stable frame that emphasizes the harmony between her motion and the tranquil setting. The fence, draped in climbing roses and wildflowers, frames the scene with natural beauty, while the weathered wooden house in the background adds a sense of timeless charm. The lighting is soft and diffused, casting gentle shadows and enhancing the nostalgic atmosphere. Her expression is serene, eyes closed at times, as if lost in the rhythm of the moment. The entire scene unfolds in a single, unbroken shot, with no cuts or shifts in perspective, allowing the viewer to fully immerse in the grace and stillness of the moment.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 67,
+ 2048
+ ],
+ "t5_context_shape": [
+ 274,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/840b214afead/bernini_receipt.json
new file mode 100644
index 0000000..b1dff5e
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/840b214afead/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"15992f2246144c6ed07475f7c9c58de924db6ab3341cf2490300c7a0a371a5b2","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"c1f45924af0cf2c17d425fc6c25b51b74cc980f4c8d8c322224835504222c9eb","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"1b62b0dbeab6dc5846d0dc29269ca1a1529b274ff336a63a6a48a7b661ab275b","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"b1dcd2064d8530cff990cb79f835d1e9bfc9825c4838dedd51d2b71d809d0bdb","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"1ec3cf0b33e9da06d58748bfccb7ad57e5b3962f4aae6f897dd17631af9cd217","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"7a5712dbc0d5f42e15bb265b557768e845070ace4f02db5f43160d56ddc3f404","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"a5da9b87e21f746d1083fa429f04a45bb8282b23c51604d5bb792195ca7a0453","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"bcd4135f5d56fae7faeb3b58b7e73aeb3f76ca0cd1e701721880fca802c4a48b","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"37cf604ec6fe8a68353bf61bde5afd00869362861576a998a59d7a833f2cf680","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"7d6e92cc7bc79a1be341ad324d892b3e8fe01c702a8b644e0363e06aab11613f","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"0251923627620990a6a4c666eb6add662a1eaf1501456093b2dc4f019a39929e","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"965868b1b4505e33716bfe157304bc05a8c743bad5c8dc02470591bf7768b183","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"11c5947118886cebcd1bf6a1eecc823a8e13ad5a438b951b2f32934fd19522b7","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"2a5cf9984339fa101b5f3291d47ee64a270aef944d64847014093d2f3efe994a","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"0a4ae2cab750190cf746b7839bddc4cbe8a1d5b92ebe0bdc1a7c92bb652ed0b7","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"20af8631c22b9d5183c253ed91205439eeb51eb5852477000f37d0f4a3448575","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"0a4ae2cab750190cf746b7839bddc4cbe8a1d5b92ebe0bdc1a7c92bb652ed0b7","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081913,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"20af8631c22b9d5183c253ed91205439eeb51eb5852477000f37d0f4a3448575","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"682d187e508391191be118335d745986443f758560808935124d41e27192b2ee","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"20af8631c22b9d5183c253ed91205439eeb51eb5852477000f37d0f4a3448575"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"0a4ae2cab750190cf746b7839bddc4cbe8a1d5b92ebe0bdc1a7c92bb652ed0b7","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"20af8631c22b9d5183c253ed91205439eeb51eb5852477000f37d0f4a3448575","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"0a4ae2cab750190cf746b7839bddc4cbe8a1d5b92ebe0bdc1a7c92bb652ed0b7","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081913,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"20af8631c22b9d5183c253ed91205439eeb51eb5852477000f37d0f4a3448575","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"60e97ee622b93c52a8c7eb6085f0faf59e880ca7d44992f32940d7735dd812cf","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"20af8631c22b9d5183c253ed91205439eeb51eb5852477000f37d0f4a3448575"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":370,"action_prompt_utf8_sha256":"effdf094385a4f2486391efc008150b7436a8137c1d5766864a678ed6e0c749f","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/840b214afead/source-exact81.mp4","source_video_sha256":"a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"1689f4723a2694b5e29c4912c91268fd8a16757433187e2d4cde44bc531f9e8f","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/rv2v.mp4","sha256":"805f6a2211b7a76e16cd85f508b341f2f06db3a04eaa336e688fe098bf5b545e","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"3c6007e70156d84f666cd1cf0a818a74cdb9c04afece892e59569de0e5f8edd7","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/t2v.mp4","sha256":"412399051be25c19ec9ab7d1406b1e6982e31e75cfa1b3920259a6c22f66113b","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646","original_pre_snapshot_sha256":"a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"a6e42a447d7fef26b073878c551a631afd6371987910a332e51e4a9e4dfb4646","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"604bff69e9f43990de2efd7c26e64d15b4f1e92d9c165d182c7e2707f9299251","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"a267d42ed5d6b8c080308914f0062fd01d9e5ac525637ff8aff050de28e1b298","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"e4d14f7d37fa5c7618f0971d9dd26ba3893976d748054ab0d688f290b3ce9dd0","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081913,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081913,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"5297c69ee8399f59a973efc144cd7ec61b8f2eb6c41677279ebb707a227836c7","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/840b214afead/bernini_t2v.jpg
new file mode 100644
index 0000000..4e9d8e6
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/840b214afead/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/840b214afead/bernini_t2v.mp4
new file mode 100644
index 0000000..30363a0
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/840b214afead/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/launch.json b/projects/video-action-editing/omni-t2v12/results/840b214afead/launch.json
new file mode 100644
index 0000000..d855a4c
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/840b214afead/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 4,
+ "case_id": "840b214afead",
+ "name": "Place bottle on treadmill",
+ "page": 7,
+ "seed": 2026081913,
+ "prompt": "A young woman with long reddish-blonde hair, wearing a mint green athletic outfit, stands on a treadmill in a modern gym with blue and white striped pillars, holding a water bottle. In this continuous shot, The woman turns her head to the left and places her water bottle onto the treadmill. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "effdf094385a4f2486391efc008150b7436a8137c1d5766864a678ed6e0c749f",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/t2v.mp4",
+ "bernini_t2v_sha256": "412399051be25c19ec9ab7d1406b1e6982e31e75cfa1b3920259a6c22f66113b",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/840b214afead/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081913",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/840b214afead/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "860178e2c937fc735b0a872c3430750f72c3801c03c53a67d62833e8e5f7be26",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:19:05.594446+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081913
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/840b214afead/omni_t2v.jpg
new file mode 100644
index 0000000..53bb951
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/840b214afead/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/840b214afead/omni_t2v.mp4
new file mode 100644
index 0000000..0372150
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/840b214afead/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/840b214afead/real_target.jpg
new file mode 100644
index 0000000..adf5f76
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/840b214afead/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/840b214afead/real_target.mp4
new file mode 100644
index 0000000..6ed4613
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/840b214afead/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/840b214afead/real_target_receipt.json
new file mode 100644
index 0000000..efd21c1
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/840b214afead/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "840b214afead",
+ "collected_utc": "2026-09-23T17:35:06.913355+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/a22aa389-eba0-4b78-815f-2e0f7956e226-00.00.02.420-00.00.04.312-seg2.mp4",
+ "sha256": "f0de03ba9c925175c5ea397a8e75203d6e4d31bed3b2baf249122f4dc0d24c3c"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "25/1",
+ "duration": "1.880000",
+ "nb_read_frames": "47"
+ },
+ "display": {
+ "path": "results/840b214afead/real_target.mp4",
+ "sha256": "989454c22b5f6db63ebd8667c943bb2622cfc2c25990661805dda3e1f9fde52b",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/840b214afead.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*1.72340425531915,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/840b214afead/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/receipt.json b/projects/video-action-editing/omni-t2v12/results/840b214afead/receipt.json
new file mode 100644
index 0000000..1c2c53a
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/840b214afead/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 4,
+ "case_id": "840b214afead",
+ "name": "Place bottle on treadmill",
+ "page": 7,
+ "seed": 2026081913,
+ "prompt": "A young woman with long reddish-blonde hair, wearing a mint green athletic outfit, stands on a treadmill in a modern gym with blue and white striped pillars, holding a water bottle. In this continuous shot, The woman turns her head to the left and places her water bottle onto the treadmill. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "effdf094385a4f2486391efc008150b7436a8137c1d5766864a678ed6e0c749f",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/t2v.mp4",
+ "bernini_t2v_sha256": "412399051be25c19ec9ab7d1406b1e6982e31e75cfa1b3920259a6c22f66113b",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/840b214afead/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/840b214afead/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081913",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/840b214afead/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "860178e2c937fc735b0a872c3430750f72c3801c03c53a67d62833e8e5f7be26",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:19:05.594446+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081913
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A young woman with long reddish-blonde hair, wearing a mint green athletic outfit, stands on a treadmill in a modern gym with blue and white striped pillars, holding a water bottle. In this continuous shot, The woman turns her head to the left and places her water bottle onto the treadmill. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young woman with long, sun-kissed reddish-blonde hair stands centered on a treadmill in a sleek, modern gym, her posture relaxed yet alert. She wears a form-fitting mint green athletic outfit—a moisture-wicking tank top and matching leggings—that contrasts softly against the cool-toned environment. The gym is illuminated by bright, even lighting that highlights the clean lines of the space, with blue and white striped pillars rising in the background, creating a rhythmic visual pattern that adds depth to the scene. The camera holds a steady, medium-wide shot, capturing her from the waist up, with the treadmill’s control panel and running belt clearly visible. In a smooth, continuous motion, she turns her head slightly to the left, her gaze briefly shifting toward the side of the treadmill, then lowers her right hand to place a clear water bottle with a blue cap onto the treadmill’s console. The bottle rests securely, and she remains in place, her expression calm and focused, as the scene continues without interruption. The lighting remains consistent, the framing stable, and the background elements—pillars, mirrors, and distant gym equipment—remain unchanged, emphasizing the quiet, deliberate nature of the moment.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 82,
+ 2048
+ ],
+ "t5_context_shape": [
+ 289,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 385
+ ],
+ "rank_zero_log_sha256": "d865c4c2b75642465de5471b399e7825250690c84f63f207fafa7208dc82ca6d"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/840b214afead/omni_t2v.mp4",
+ "sha256": "f8c6c74bac6feaeb53623326afaf4ba7a839d47b75d8981bd400429fd9cd5cd1",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 555.63,
+ "completed_utc": "2026-09-23T08:28:03.800437+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/840b214afead/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/840b214afead/t2v_conditioning.json
new file mode 100644
index 0000000..2c32597
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/840b214afead/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A young woman with long reddish-blonde hair, wearing a mint green athletic outfit, stands on a treadmill in a modern gym with blue and white striped pillars, holding a water bottle. In this continuous shot, The woman turns her head to the left and places her water bottle onto the treadmill. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young woman with long, sun-kissed reddish-blonde hair stands centered on a treadmill in a sleek, modern gym, her posture relaxed yet alert. She wears a form-fitting mint green athletic outfit\u2014a moisture-wicking tank top and matching leggings\u2014that contrasts softly against the cool-toned environment. The gym is illuminated by bright, even lighting that highlights the clean lines of the space, with blue and white striped pillars rising in the background, creating a rhythmic visual pattern that adds depth to the scene. The camera holds a steady, medium-wide shot, capturing her from the waist up, with the treadmill\u2019s control panel and running belt clearly visible. In a smooth, continuous motion, she turns her head slightly to the left, her gaze briefly shifting toward the side of the treadmill, then lowers her right hand to place a clear water bottle with a blue cap onto the treadmill\u2019s console. The bottle rests securely, and she remains in place, her expression calm and focused, as the scene continues without interruption. The lighting remains consistent, the framing stable, and the background elements\u2014pillars, mirrors, and distant gym equipment\u2014remain unchanged, emphasizing the quiet, deliberate nature of the moment.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 82,
+ 2048
+ ],
+ "t5_context_shape": [
+ 289,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/bernini_receipt.json
new file mode 100644
index 0000000..f340e8a
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"0e3ef520bc12f16c3322dfeb5810d463e7bee5e728de439339b38512bb141137","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"cd29b62f704ca8c1579cb3a0c767346d28b684484398cfc971b948296f5aed07","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"0fa7f7a5b411878dd2248e630ed1a3d7477e4bb0460afe7949252f2694aef383","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"8ffb30e2874b54196df648959b1669d00a4d5f7daf821926456b1699eb3b7f46","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"43b69b00ada967ef76345826b0674fe7d841f244ea90523b88e9f166cb437d08","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"316d1888f0ee198f627be953524eb1c7aba5f4da4abacf123c2e6a9b44905766","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"75d93e68bf8a1d89bba7cb0f91958c8300d371794698ab96f7949dc8877007eb","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"9222c48b1aa81b3e05f3fce8d36674e040c8cc68eec751848393d9bc193e19ae","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"40e2b7b43ba9dbb148d76b5caccf57364ba283aa1fad74534466b06ecaef0858","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"5719ed9082ae985b7a8c72202e49d83c541621ce1571331b9193cedfd25fdd3b","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"b392d25ca7581c2810be2badac1880ea655b8d9e9ad28690a62348ee65d16784","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"dc7e43abdf0bfc78fd327747c6dba38e7a1c2b74bc1ea2db7ff7544bb3a14a24","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"833c04baee1212aea75bc6325dca38b7761ba030a5c5dc52ff8d4e02f1eae3c7","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"e1a0535630291d954cd87b1c9f8ef201c5a66512479252f810849031753aac72","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"dad5739e74c429c4e4df7fc1727082b13a2bed19e9edd37cf798a5b223207324","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"60c52f35fe1da3abe023e4bd9124bbe22bf38d9b2d916961e42fe20febf9e8c3","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"dad5739e74c429c4e4df7fc1727082b13a2bed19e9edd37cf798a5b223207324","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081907,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"60c52f35fe1da3abe023e4bd9124bbe22bf38d9b2d916961e42fe20febf9e8c3","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"165f816820831627f57ef91cb2359113994527313b0325cc45e2306c78011673","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"60c52f35fe1da3abe023e4bd9124bbe22bf38d9b2d916961e42fe20febf9e8c3"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"dad5739e74c429c4e4df7fc1727082b13a2bed19e9edd37cf798a5b223207324","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"60c52f35fe1da3abe023e4bd9124bbe22bf38d9b2d916961e42fe20febf9e8c3","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"dad5739e74c429c4e4df7fc1727082b13a2bed19e9edd37cf798a5b223207324","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081907,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"60c52f35fe1da3abe023e4bd9124bbe22bf38d9b2d916961e42fe20febf9e8c3","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"9111a8c3be8ce6f71847e1dd77c43d5f155206271b60b94e91215602234e2133","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"60c52f35fe1da3abe023e4bd9124bbe22bf38d9b2d916961e42fe20febf9e8c3"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":293,"action_prompt_utf8_sha256":"3c320866e1c96c3a128a525a9d0e4633d6cb1aef7b519e9c623c841f3f004e31","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d5221a1b69d5/source-exact81.mp4","source_video_sha256":"fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"4e2b456b1820d80d383681df6cb74752dd961cea36fb8bb4fc6e82ad46a75043","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/rv2v.mp4","sha256":"5cb52ddd029742d7fd8d7b6e6ef04df46dd6d3fb77a3522b4bd8f36f5afde38e","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"1fc6f50f9c555be06be5c88d710adb9ee2fb644f24875d703c3df5b877f72ee3","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/t2v.mp4","sha256":"c721c58707f97a876842d467b7e3e5ebe84eb41021512af9bdfa33ac837717d2","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14","original_pre_snapshot_sha256":"fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"fdd1b3df4b94449dc59cd04d03deb769fb2cf6c4c2d7b87221e2b3a6b5557d14","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"88083a3f93b4d2444e7e101a6aca51f275f1c3263bec08f076053272d05c7a27","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"f01ab3d32bb9c931c4b5f2ccc903f8e5a50d2a5a8bd783287f4b115838dca51c","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"409d0f5ee949db66d90b9b3a322400df21acb20a42df2d668b401bcfad6e8e6f","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081907,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081907,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"5252b29687addade7b735cfb6dfcebac9942f72444bfa98d739981c82f036fc3","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/bernini_t2v.jpg
new file mode 100644
index 0000000..dc97d2e
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/bernini_t2v.mp4
new file mode 100644
index 0000000..e71c2d2
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/launch.json b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/launch.json
new file mode 100644
index 0000000..a14b7a9
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 10,
+ "case_id": "d5221a1b69d5",
+ "name": "闭眼尝汤 · Taste from ladle",
+ "page": null,
+ "seed": 2026081907,
+ "prompt": "A young woman in a patterned apron stands in a modern kitchen, tasting food from a ladle with a satisfied expression. In this continuous shot, The woman in an apron closes her eyes as she tastes from a metal ladle. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "3c320866e1c96c3a128a525a9d0e4633d6cb1aef7b519e9c623c841f3f004e31",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/t2v.mp4",
+ "bernini_t2v_sha256": "c721c58707f97a876842d467b7e3e5ebe84eb41021512af9bdfa33ac837717d2",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d5221a1b69d5/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081907",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d5221a1b69d5/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "c05aa2bfd6d9ab38473e6b667f8fb640b0d6b1126adfb05f9a9a74a8cab7a553",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:19:38.654569+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081907
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/omni_t2v.jpg
new file mode 100644
index 0000000..b1f08af
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/omni_t2v.mp4
new file mode 100644
index 0000000..908a6d6
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/real_target.jpg
new file mode 100644
index 0000000..c6a16fa
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/real_target.mp4
new file mode 100644
index 0000000..0abad7d
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/real_target_receipt.json
new file mode 100644
index 0000000..b779ff5
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "d5221a1b69d5",
+ "collected_utc": "2026-09-23T17:35:23.118612+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/6a7753fa-f254-4570-8369-918954606350-00.00.03.841-00.00.07.536-seg2.mp4",
+ "sha256": "5da5df9d958b4f56a5b5b545e059ac9288471e27a0e8b36fac2c5a92a4f63b77"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "25/1",
+ "duration": "3.680000",
+ "nb_read_frames": "92"
+ },
+ "display": {
+ "path": "results/d5221a1b69d5/real_target.mp4",
+ "sha256": "3b61c8123a193d9feab392a7780ea5b87b91099a2c41ea069875febc81ea9f4e",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/d5221a1b69d5.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*0.880434782608696,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/d5221a1b69d5/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/receipt.json b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/receipt.json
new file mode 100644
index 0000000..746e8a1
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 10,
+ "case_id": "d5221a1b69d5",
+ "name": "闭眼尝汤 · Taste from ladle",
+ "page": null,
+ "seed": 2026081907,
+ "prompt": "A young woman in a patterned apron stands in a modern kitchen, tasting food from a ladle with a satisfied expression. In this continuous shot, The woman in an apron closes her eyes as she tastes from a metal ladle. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "3c320866e1c96c3a128a525a9d0e4633d6cb1aef7b519e9c623c841f3f004e31",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/t2v.mp4",
+ "bernini_t2v_sha256": "c721c58707f97a876842d467b7e3e5ebe84eb41021512af9bdfa33ac837717d2",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d5221a1b69d5/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d5221a1b69d5/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081907",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d5221a1b69d5/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "c05aa2bfd6d9ab38473e6b667f8fb640b0d6b1126adfb05f9a9a74a8cab7a553",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:19:38.654569+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081907
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A young woman in a patterned apron stands in a modern kitchen, tasting food from a ladle with a satisfied expression. In this continuous shot, The woman in an apron closes her eyes as she tastes from a metal ladle. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young woman stands in a bright, modern kitchen, her figure centered in a steady, continuous shot. She wears a patterned apron over casual clothing, the fabric soft and slightly rumpled from use. The kitchen is sleek and well-lit, with natural light streaming through a large window, casting a warm glow across stainless steel appliances and clean white countertops. In her hand, she holds a metal ladle, its surface gleaming under the light. She lifts it slowly to her lips, her eyes closing gently as she tastes the food. Her expression softens into quiet satisfaction—lips parting slightly, a subtle smile forming as the flavor registers. The camera remains fixed, capturing the moment with unwavering focus, preserving the stillness of the scene. The lighting remains consistent, highlighting the contours of her face and the texture of the apron. The background remains unchanged—neatly arranged utensils, a pot simmering on the stove, a bowl of fresh herbs on the counter—all contributing to the calm, domestic atmosphere. The shot lingers, emphasizing the sensory experience, the quiet joy of a meal well-prepared, and the simple pleasure of tasting one’s own creation.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 70,
+ 2048
+ ],
+ "t5_context_shape": [
+ 274,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 358
+ ],
+ "rank_zero_log_sha256": "553659a0edaf6f48af82c44481a5de73ec1c71aac3c3a88556e15fb616195053"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/d5221a1b69d5/omni_t2v.mp4",
+ "sha256": "77cbffd0f6650d641829a18b2f5dc7244c69ee0d430ca1f5e2a5efaa44e984e2",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 552.79,
+ "completed_utc": "2026-09-23T08:28:25.409926+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/t2v_conditioning.json
new file mode 100644
index 0000000..ce46a80
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d5221a1b69d5/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A young woman in a patterned apron stands in a modern kitchen, tasting food from a ladle with a satisfied expression. In this continuous shot, The woman in an apron closes her eyes as she tastes from a metal ladle. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young woman stands in a bright, modern kitchen, her figure centered in a steady, continuous shot. She wears a patterned apron over casual clothing, the fabric soft and slightly rumpled from use. The kitchen is sleek and well-lit, with natural light streaming through a large window, casting a warm glow across stainless steel appliances and clean white countertops. In her hand, she holds a metal ladle, its surface gleaming under the light. She lifts it slowly to her lips, her eyes closing gently as she tastes the food. Her expression softens into quiet satisfaction\u2014lips parting slightly, a subtle smile forming as the flavor registers. The camera remains fixed, capturing the moment with unwavering focus, preserving the stillness of the scene. The lighting remains consistent, highlighting the contours of her face and the texture of the apron. The background remains unchanged\u2014neatly arranged utensils, a pot simmering on the stove, a bowl of fresh herbs on the counter\u2014all contributing to the calm, domestic atmosphere. The shot lingers, emphasizing the sensory experience, the quiet joy of a meal well-prepared, and the simple pleasure of tasting one\u2019s own creation.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 70,
+ 2048
+ ],
+ "t5_context_shape": [
+ 274,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/d908552fb691/bernini_receipt.json
new file mode 100644
index 0000000..a394030
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d908552fb691/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"a0a093221006a4bc061d2c49d193ac3553aef2db2c994a5121d3953e7ec9f50c","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"60f4c1f2a8fa41430dfd460cf678c517c8d2a0864c6aae9931af4e99f24cc24a","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"08ac76711877a7c3c995081b07d21cab4e28dde1768dc4bd06293d2d0e594dd8","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"5afd39a16b79c04f11b123c7e732891200577346c56c6ff34194bdffca79ac7d","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"67e8cf261d8cb4f8842d5848348ad6d157a2d135fce8018f403566a68c18103c","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"e3bf8e07296b0cdcf862746603667335e86050993c53b494f51f296267237f0b","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"9285004abd79644620eb8f144c051849bb29ce584b4b4483a594456386a03e3b","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"785548ee49b875d2fcfab42cd65efdfb3516a7985c777051e8557c80580b3e56","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"400d2e2c096a5ba77d2632fa77fe0dec3e1c092e16c534a6257f23476e94c352","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"b1b95705d274ce5a92933877856d9e703f8cf0ccbda9178b0ca1ee1b382a2b17","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"9896e66b548f5147ecb12cf96e88f128c27406e6dec8350fb5972f7b7e935368","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"3849e04624c5c84e6733c1ce7edad70384ee977aa138f6e143d94e45f978400f","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"92549679f4e010e231e5b2dcdda92736c20466762730683cea7f8932fc4eb97c","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"64eb180444ab66420f959ae3a668ad3e51a80abbab6c7384bd98c48e25100270","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"6de6325eb03ecd284f77eabb2a050a89730d7f0463ef36f7ca728782897a36a5","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"e3477e803918b5d84ff200b2be049c36f69a18c79419595cc7f18e3684b2246e","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"6de6325eb03ecd284f77eabb2a050a89730d7f0463ef36f7ca728782897a36a5","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081903,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"e3477e803918b5d84ff200b2be049c36f69a18c79419595cc7f18e3684b2246e","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"e5564515e89d41c7277fa82415b91e753bd875abafe5501dfec22bbc2d0f6c36","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"e3477e803918b5d84ff200b2be049c36f69a18c79419595cc7f18e3684b2246e"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"6de6325eb03ecd284f77eabb2a050a89730d7f0463ef36f7ca728782897a36a5","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"e3477e803918b5d84ff200b2be049c36f69a18c79419595cc7f18e3684b2246e","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"6de6325eb03ecd284f77eabb2a050a89730d7f0463ef36f7ca728782897a36a5","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081903,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"e3477e803918b5d84ff200b2be049c36f69a18c79419595cc7f18e3684b2246e","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"8018b358ab8ea9e4cb9bcb228505076ad7ab084bb85c5d0706278b7073345b71","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"e3477e803918b5d84ff200b2be049c36f69a18c79419595cc7f18e3684b2246e"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":382,"action_prompt_utf8_sha256":"6c693b009c7e8c0d2f3c4ed680d119e2ef14e800cbabc6a10da0ff2556bb86b8","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d908552fb691/source-exact81.mp4","source_video_sha256":"56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"eed978266371e00b3feb9b7d17dcc61633f46d5ea90722e2e3e19ab61090be37","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/rv2v.mp4","sha256":"2ab7621e74b848a998822edb340ce07baca308814bb5524d7d3cd910e45e9efd","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"c1b0f92804195de0b39a900d62bf435141f4a8a095cb56abf4dee562742b940f","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/t2v.mp4","sha256":"704f866e6b6a0615f87d62f9815c61c065519e900d2e70be5a517513994daa3d","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb","original_pre_snapshot_sha256":"56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"56e1faa78c02c32d6da71354e8434241c8e9a4c334da4ac8aa4415148b6230bb","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"5f2cf4cedb92a86ab416ff83ba501d1e2e6647dd9d86fd5cd2e68ac729dbe68d","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"e8b542250790de1613b15c355501c22efa624817869e7528d8a09ef141a21a1f","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"c3f1aae496efdb016048dfc552989cc37daa8b7724258214b7a1e81adaa1c2a0","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081903,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081903,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"ec20637b3b2070c1949c0f06d95e6d6717facc1dedc3c4d52aea479e96b281a7","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/d908552fb691/bernini_t2v.jpg
new file mode 100644
index 0000000..8766b07
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d908552fb691/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/d908552fb691/bernini_t2v.mp4
new file mode 100644
index 0000000..884df3d
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d908552fb691/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/launch.json b/projects/video-action-editing/omni-t2v12/results/d908552fb691/launch.json
new file mode 100644
index 0000000..6691ef5
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d908552fb691/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 5,
+ "case_id": "d908552fb691",
+ "name": "Bow to clap",
+ "page": 7,
+ "seed": 2026081903,
+ "prompt": "A woman in a white taekwondo uniform with a black collar and Adidas logo stands in profile against a black background under dramatic side lighting, captured in a static medium shot. In this continuous shot, The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "6c693b009c7e8c0d2f3c4ed680d119e2ef14e800cbabc6a10da0ff2556bb86b8",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/t2v.mp4",
+ "bernini_t2v_sha256": "704f866e6b6a0615f87d62f9815c61c065519e900d2e70be5a517513994daa3d",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d908552fb691/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081903",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d908552fb691/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "9f11a79adac35975bf5bcb465e3c8e935472854124d0997dc6922841bc903c30",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:28:21.078158+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081903
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/d908552fb691/omni_t2v.jpg
new file mode 100644
index 0000000..a72a488
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d908552fb691/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/d908552fb691/omni_t2v.mp4
new file mode 100644
index 0000000..0709854
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d908552fb691/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/d908552fb691/real_target.jpg
new file mode 100644
index 0000000..e18fd6c
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d908552fb691/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/d908552fb691/real_target.mp4
new file mode 100644
index 0000000..54a9897
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d908552fb691/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/d908552fb691/real_target_receipt.json
new file mode 100644
index 0000000..766a0e3
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d908552fb691/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "d908552fb691",
+ "collected_utc": "2026-09-23T17:35:06.549658+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/317a8639-705d-42b8-b51a-7226e7a3c7a6-00.00.06.631-00.00.10.787-seg2.mp4",
+ "sha256": "935406db5a168206aea061e44f9f577135605454e1ccb0eaea91828ec28f2d0f"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "24000/1001",
+ "duration": "4.170833",
+ "nb_read_frames": "100"
+ },
+ "display": {
+ "path": "results/d908552fb691/real_target.mp4",
+ "sha256": "ff161cba09e00e281c6407a3e96e7e359f9900a4d0af4dc6a0d042931aeadcf6",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/d908552fb691.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*0.776823238906952,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/d908552fb691/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/receipt.json b/projects/video-action-editing/omni-t2v12/results/d908552fb691/receipt.json
new file mode 100644
index 0000000..06c0c71
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d908552fb691/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 5,
+ "case_id": "d908552fb691",
+ "name": "Bow to clap",
+ "page": 7,
+ "seed": 2026081903,
+ "prompt": "A woman in a white taekwondo uniform with a black collar and Adidas logo stands in profile against a black background under dramatic side lighting, captured in a static medium shot. In this continuous shot, The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "6c693b009c7e8c0d2f3c4ed680d119e2ef14e800cbabc6a10da0ff2556bb86b8",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/t2v.mp4",
+ "bernini_t2v_sha256": "704f866e6b6a0615f87d62f9815c61c065519e900d2e70be5a517513994daa3d",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d908552fb691/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d908552fb691/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081903",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d908552fb691/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "9f11a79adac35975bf5bcb465e3c8e935472854124d0997dc6922841bc903c30",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:28:21.078158+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081903
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A woman in a white taekwondo uniform with a black collar and Adidas logo stands in profile against a black background under dramatic side lighting, captured in a static medium shot. In this continuous shot, The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A woman in a crisp white taekwondo uniform stands in profile against a deep black background, her form sharply defined by dramatic side lighting that casts a strong, sculpted shadow along her silhouette. The uniform features a black collar and a subtle Adidas logo on the chest, adding a touch of brand detail to the athletic attire. She is captured in a static medium shot, framed with precision to highlight her poised stance and the clean lines of her posture. The lighting emphasizes the contours of her body, creating a high-contrast, cinematic effect that enhances the intensity of the moment. In a smooth, continuous motion, she brings her hands together in front of her chest, clapping them once with a deliberate, controlled gesture. Her movements are precise and measured, reflecting discipline and focus. The scene remains unchanged throughout—no shifts in camera angle, lighting, or background—maintaining a consistent, composed atmosphere that centers entirely on her presence and action. The stillness of the frame contrasts with the subtle energy of her motion, creating a powerful visual statement of control and readiness.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 85,
+ 2048
+ ],
+ "t5_context_shape": [
+ 236,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 335
+ ],
+ "rank_zero_log_sha256": "4f029607a0c836fa49e9a318191fdfebd6a1aff88df0fe004ff5b9ba44aafa00"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/d908552fb691/omni_t2v.mp4",
+ "sha256": "97ca0d377f030e084df41b86b49ad784496e336a90223c1acd81eb886f8366f6",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 533.44,
+ "completed_utc": "2026-09-23T08:36:57.331581+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/d908552fb691/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/d908552fb691/t2v_conditioning.json
new file mode 100644
index 0000000..574a153
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d908552fb691/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A woman in a white taekwondo uniform with a black collar and Adidas logo stands in profile against a black background under dramatic side lighting, captured in a static medium shot. In this continuous shot, The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A woman in a crisp white taekwondo uniform stands in profile against a deep black background, her form sharply defined by dramatic side lighting that casts a strong, sculpted shadow along her silhouette. The uniform features a black collar and a subtle Adidas logo on the chest, adding a touch of brand detail to the athletic attire. She is captured in a static medium shot, framed with precision to highlight her poised stance and the clean lines of her posture. The lighting emphasizes the contours of her body, creating a high-contrast, cinematic effect that enhances the intensity of the moment. In a smooth, continuous motion, she brings her hands together in front of her chest, clapping them once with a deliberate, controlled gesture. Her movements are precise and measured, reflecting discipline and focus. The scene remains unchanged throughout\u2014no shifts in camera angle, lighting, or background\u2014maintaining a consistent, composed atmosphere that centers entirely on her presence and action. The stillness of the frame contrasts with the subtle energy of her motion, creating a powerful visual statement of control and readiness.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 85,
+ 2048
+ ],
+ "t5_context_shape": [
+ 236,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/bernini_receipt.json
new file mode 100644
index 0000000..e2bcd41
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"72c86e50643180d5462a1cfbd7866503bd47e27b48016cca61e78cbf494287cd","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"df07812ea8d48d3e0fa0ab394ed62e096dd60fcfdfd28616360b60173561fb58","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"bd952f750ac9c0fba0da81749ba7954bb9f023125ce0d7f94e6528a7aec3b972","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"1821d12e74e20c62a3406da3c39354e2e1dff719acc674fe03e1ea3d5ed3039f","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"ad029d8836050c90ebbdeac146df913676f82f1a957d13cbba7fc274dddae653","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"94bab1c3e870a44bb147ab021ba86dd13fe79a9d9236e96e21bca3029f3b9eec","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"a8ac5e4820c0b58cb6b544496076ce1c54641298dd341f135cfb575c764e225f","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"e63ae6bdcc2cd4fc6a1c6c09877fb70a792e194da858195f4adcd9865e516365","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"fc5dceb5bfc6a8a50aa23d4e2e6cc15d6d558172c3e34a85650f37f611674eb5","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"1a56ce494d1b21394b3ad3d7b75d3592116a4b7599f14b99bebb1f8b5e41ff35","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"d110645f18034ba60a6e4d7414c132abdfcb76fb788804ec612af60fc3c5bddf","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"d25a21c01fd7b4e758ef172da8d4f953a0106d5e676d0c2cc6c2aed17c921c02","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"e25ae302ef64edd9f14a03fef40e678245d7e6bd8c3e785d23d69cc319ce3fb6","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"865adedaf9f9db1049576610e242cde606d7439dea343619f73c82d8b7ca450f","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"e55f92c066aea4ceb4049375ab53d391e12a4cc8fef0bd2268e7409663da7540","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"01d9f9d890d22ab6a60aadd0143efb87aab45160079a3f4660b3856b0cb3b064","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"e55f92c066aea4ceb4049375ab53d391e12a4cc8fef0bd2268e7409663da7540","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081905,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"01d9f9d890d22ab6a60aadd0143efb87aab45160079a3f4660b3856b0cb3b064","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"d065f9f4a2e85a3097303bfa6832615de7d2abe241636716db2208f7083dc05b","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"01d9f9d890d22ab6a60aadd0143efb87aab45160079a3f4660b3856b0cb3b064"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"e55f92c066aea4ceb4049375ab53d391e12a4cc8fef0bd2268e7409663da7540","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"01d9f9d890d22ab6a60aadd0143efb87aab45160079a3f4660b3856b0cb3b064","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"e55f92c066aea4ceb4049375ab53d391e12a4cc8fef0bd2268e7409663da7540","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081905,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"01d9f9d890d22ab6a60aadd0143efb87aab45160079a3f4660b3856b0cb3b064","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"828064c90e3e07930cbcec5a0fba7d48bb6d3cb493e763f07fffda4724696e27","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"01d9f9d890d22ab6a60aadd0143efb87aab45160079a3f4660b3856b0cb3b064"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":319,"action_prompt_utf8_sha256":"e405c3746dde5e6718d5d165841db90f2078d73391a04c4e6d309c75e59cea3a","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/d977ce147d80/source-exact81.mp4","source_video_sha256":"e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"9d707f3f3e2c15e1ad8fadf35169538f7e380628620773448cc85a9b25dd5058","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/rv2v.mp4","sha256":"ca8c8d8162467af99f51cac705c13ddbdf6a30c19f2b4d75053c9ad71f2a44cd","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"1b387a9a30058c171af5dbbf67722c8edfbeda277f13ed7cf10012f10743d77b","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/t2v.mp4","sha256":"baac904b0402dfab39c956835d36f5c37e2a079d59d47517857c702f3f42f5d6","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d","original_pre_snapshot_sha256":"e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"e2a056e1af423be25a5beb4b75d1145fbacd625e3316a8e64b6031e75739f80d","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"9cac90cd9e7113aedd3a75fbb3867e553dac93dd7f0077b11d0328635cc90f25","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"b2cb6032bd7dd832c9c87724807bf4206caff68c9a84183460e721087847b768","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"77a92dd70655be7c3a80d59899974e3c0b9bfaa7e1a64b170e51f1b5ccc1a722","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081905,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081905,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"0624a1d4c2f2e1d6d96c4739535d98833a1914de2ae6d6984ed694278ac27afd","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/bernini_t2v.jpg
new file mode 100644
index 0000000..256df07
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/bernini_t2v.mp4
new file mode 100644
index 0000000..53e617d
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/launch.json b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/launch.json
new file mode 100644
index 0000000..3500590
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 3,
+ "case_id": "d977ce147d80",
+ "name": "Rub eyes then cover face",
+ "page": 6,
+ "seed": 2026081905,
+ "prompt": "A young man with dark skin, dreadlocks, and a gold septum ring lies on a dark couch in a dimly lit room, showing signs of distress. In this continuous shot, The man lying on the couch rubs his tired eyes and covers his face with both hands. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "e405c3746dde5e6718d5d165841db90f2078d73391a04c4e6d309c75e59cea3a",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/t2v.mp4",
+ "bernini_t2v_sha256": "baac904b0402dfab39c956835d36f5c37e2a079d59d47517857c702f3f42f5d6",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d977ce147d80/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081905",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d977ce147d80/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "54e178e1d21b606a4d9ae16c9a4df2e90b15e7f14afee9b6deeb8f5d88ac797b",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:06:32.200611+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081905
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/omni_t2v.jpg
new file mode 100644
index 0000000..73144f6
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/omni_t2v.mp4
new file mode 100644
index 0000000..7e6d5c5
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/real_target.jpg
new file mode 100644
index 0000000..b736e2e
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/real_target.mp4
new file mode 100644
index 0000000..de93c0d
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/real_target_receipt.json
new file mode 100644
index 0000000..fce1b85
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "d977ce147d80",
+ "collected_utc": "2026-09-23T17:35:05.571840+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/4867ced3-df82-498c-876b-4452fe7ffec1-00.00.08.489-00.00.12.425-seg3.mp4",
+ "sha256": "ac544e4c77f2fa0650d9397cd1fafaff40e6b3ac673385c69cf723fc754ebd30"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "24/1",
+ "duration": "3.916667",
+ "nb_read_frames": "94"
+ },
+ "display": {
+ "path": "results/d977ce147d80/real_target.mp4",
+ "sha256": "9a2ad5f5fc5fca2da6532ecddbf6d5879e9c09cfcb57b952b8b28758d628cb1e",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/d977ce147d80.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*0.8272339721503,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/d977ce147d80/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/receipt.json b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/receipt.json
new file mode 100644
index 0000000..96899f2
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 3,
+ "case_id": "d977ce147d80",
+ "name": "Rub eyes then cover face",
+ "page": 6,
+ "seed": 2026081905,
+ "prompt": "A young man with dark skin, dreadlocks, and a gold septum ring lies on a dark couch in a dimly lit room, showing signs of distress. In this continuous shot, The man lying on the couch rubs his tired eyes and covers his face with both hands. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "e405c3746dde5e6718d5d165841db90f2078d73391a04c4e6d309c75e59cea3a",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/t2v.mp4",
+ "bernini_t2v_sha256": "baac904b0402dfab39c956835d36f5c37e2a079d59d47517857c702f3f42f5d6",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/d977ce147d80/receipt.json",
+ "comparison_batch": "progress2_omni6_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d977ce147d80/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081905",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/d977ce147d80/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "54e178e1d21b606a4d9ae16c9a4df2e90b15e7f14afee9b6deeb8f5d88ac797b",
+ "hostname": "auh7-1b-gpu-253",
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "started_utc": "2026-09-23T08:06:32.200611+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081905
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A young man with dark skin, dreadlocks, and a gold septum ring lies on a dark couch in a dimly lit room, showing signs of distress. In this continuous shot, The man lying on the couch rubs his tired eyes and covers his face with both hands. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young man with dark skin and long dreadlocks lies sprawled across a dark, plush couch in a dimly lit room, the low ambient light casting deep shadows that pool around him. A gold septum ring glints faintly as he shifts slightly, his body tense with quiet unease. The room is still and intimate, the atmosphere heavy with a sense of emotional weight. He lifts his hands slowly, rubbing his tired eyes with the heels of his palms, his fingers pressing into the sockets as if trying to push back exhaustion or pain. Then, with a slow, deliberate motion, he covers his face entirely, pressing both hands against his forehead and cheeks, his breath shallow and uneven. The camera holds steady, capturing the moment in a continuous, unbroken shot—no movement, no cut—allowing the viewer to sit with the man’s distress. The framing remains tight and consistent, focusing on his upper body and face, emphasizing the subtle tremors in his hands and the way his shoulders rise and fall with each labored breath. The lighting remains unchanged, soft and moody, highlighting the contours of his face and the texture of his hair, while the surrounding darkness absorbs all distractions, drawing attention solely to his internal struggle. The scene is quiet, almost still, yet charged with unspoken emotion.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 78,
+ 2048
+ ],
+ "t5_context_shape": [
+ 307,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 399
+ ],
+ "rank_zero_log_sha256": "553e3d91be0dff92c44c637859f9a1428b84512cd4ae3cb64beb94aafe45df43"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/d977ce147d80/omni_t2v.mp4",
+ "sha256": "511b0763dc6b87bd4ecb80dae8ba4b95442b8c424f861e2587297e9841316edd",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 778.41,
+ "completed_utc": "2026-09-23T08:18:48.058409+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/d977ce147d80/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/t2v_conditioning.json
new file mode 100644
index 0000000..d82e569
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/d977ce147d80/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A young man with dark skin, dreadlocks, and a gold septum ring lies on a dark couch in a dimly lit room, showing signs of distress. In this continuous shot, The man lying on the couch rubs his tired eyes and covers his face with both hands. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young man with dark skin and long dreadlocks lies sprawled across a dark, plush couch in a dimly lit room, the low ambient light casting deep shadows that pool around him. A gold septum ring glints faintly as he shifts slightly, his body tense with quiet unease. The room is still and intimate, the atmosphere heavy with a sense of emotional weight. He lifts his hands slowly, rubbing his tired eyes with the heels of his palms, his fingers pressing into the sockets as if trying to push back exhaustion or pain. Then, with a slow, deliberate motion, he covers his face entirely, pressing both hands against his forehead and cheeks, his breath shallow and uneven. The camera holds steady, capturing the moment in a continuous, unbroken shot\u2014no movement, no cut\u2014allowing the viewer to sit with the man\u2019s distress. The framing remains tight and consistent, focusing on his upper body and face, emphasizing the subtle tremors in his hands and the way his shoulders rise and fall with each labored breath. The lighting remains unchanged, soft and moody, highlighting the contours of his face and the texture of his hair, while the surrounding darkness absorbs all distractions, drawing attention solely to his internal struggle. The scene is quiet, almost still, yet charged with unspoken emotion.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 78,
+ 2048
+ ],
+ "t5_context_shape": [
+ 307,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/bernini_receipt.json b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/bernini_receipt.json
new file mode 100644
index 0000000..cbd6c7f
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/bernini_receipt.json
@@ -0,0 +1 @@
+{"arms":["t2v","rv2v"],"bernini_commit":"2d2b4591ac053ec25c6371b01a5a6746679e5793","bernini_inference_files":{"bernini/cli.py":"26949fbf246003403ed0cca1ec1bbb62c2099fc9740bb17ba5a1e7c86fbc0edf","bernini/io_utils.py":"233541373746f5d97e1cb3680d3c2a41d5d212b797eefb97693afa6e3ab5f30a","bernini/pipeline.py":"c6acf05c01a637d9bce69e8160eb6eb4260ff4ec798fd990de8e5aa73999ab40"},"checkpoint":{"content":{"every_file_sha256_verified":true,"manifest_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/bernini_counterfactual_identity_orbit_v5_20260808_c099c6f/runtime/source_ea900d5/methods/bernini_action_editing/audits/bernini_r13_ff4c5d4_checkpoint.sha256","manifest_sha256_computed":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","manifest_sha256_expected":"a95ac2d74fc4379134a6276355d472810ef08e3d9de79761f1244375a6fad831","verified_entries_digest":"676e6104eebee3ab1066c70f40af385346b013a3afcab8cafb06c5290994d9ba","verified_file_count":23},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit/checkpoints/Bernini-R-1.3B-Diffusers-ff4c5d4","tree_sha256":"6be0d0db0dd483daf1a843efa2b5aafc20090ad11dc0fc6ee8859bdf150635ca"},"condition_identities":{"full_source_video":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"be2ec7c843e350f5aee025bd2c422f7f0807bfb8a00ef9163443f6d2cc0eafc8","dtype":"torch.float32","finite":true,"label":"full_source_video","numel":1267392,"raw_storage_sha256":"427cc6dd1c8de70592aa52107487a169595bfe529ff3257e8e42f86e8d96b72d","shape":[1,16,21,46,82]}},"rank_zero_broadcasts":{"full_source_video":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,21,46,82]},"prebroadcast_metadata_all_rank_exact":true},"references":{"0":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"27":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"53":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true},"80":{"authoritative_rank":0,"broadcast_before_renderer":true,"contract":{"device_type":"cuda","dtype":"torch.float32","finite":true,"shape":[1,16,1,46,82]},"prebroadcast_metadata_all_rank_exact":true}}},"references":{"0":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"1ff488257f4c4913bef90d7915692a39126bba448dbfaa353b426e60b6a55101","dtype":"torch.float32","finite":true,"label":"source_reference_0","numel":60352,"raw_storage_sha256":"b1d3c7f61584ce40d24116da78f2b01d04d56e3ed2fc0047cc560eb4506589c8","shape":[1,16,1,46,82]}},"27":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"355d79080c4a1dbab48cfe565185fc7700bc41dab83d54842108afce69feeebf","dtype":"torch.float32","finite":true,"label":"source_reference_27","numel":60352,"raw_storage_sha256":"a3467158c9eb3d9b26d1649d9c30054890ab83038ce9f4cd672cbb9c43dbd7be","shape":[1,16,1,46,82]}},"53":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"3cfef1ae9cf793abd6e40329a8c85701d64ba12b99f5c0fe8581542262124a22","dtype":"torch.float32","finite":true,"label":"source_reference_53","numel":60352,"raw_storage_sha256":"6323273e4f2f031e2e9c2c3d9e921440c437ecc9fa3c26646853f7ba7d8077e7","shape":[1,16,1,46,82]}},"80":{"all_rank_exact":true,"identity":{"byte_count":241408,"content_sha256":"6d2c3d2b8d525cf1abc7a361f0a301c2d8f7ed88f1e291b35c08fa50a1f9382a","dtype":"torch.float32","finite":true,"label":"source_reference_80","numel":60352,"raw_storage_sha256":"4c1c4a145f5ba323088768417cfbdf05b2d66a9e1da94e3fc99e3b325ac32c68","shape":[1,16,1,46,82]}}}},"conditioning":{"rv2v":{"full_source_video_count":1,"reference_encoding":"independent_rgb_frame_to_wan_vae_[1,C,1,H,W]","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":4,"source_frame_indices":[0,27,53,80],"source_ids":{"conditioning_source_count":5,"max_conditioning_source_id":5,"reference_source_ids":[2,3,4,5],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[1],"within_pretrained_source_ids_1_through_5":true}},"t2v":{"full_source_video_count":0,"reference_encoding":"none","reference_from_temporal_video_latent_slice":false,"source_derived_reference_count":0,"source_frame_indices":[],"source_ids":{"conditioning_source_count":0,"max_conditioning_source_id":0,"reference_source_ids":[],"source_id_interpolation_required":false,"target_source_id":0,"video_source_ids":[],"within_pretrained_source_ids_1_through_5":true}}},"experimental_canary":true,"freeze_certificate":{"base_frozen":true,"lora_module_count":0,"trainable_parameter_elements":0,"trainable_parameter_tensors":0},"generated_identities":{"rv2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"238f4c2e3e40cbb7ac6259faa16b3ff16e38ff7cfcc176d62e94851846de9b12","dtype":"torch.float32","finite":true,"label":"generated_rv2v","numel":1267392,"raw_storage_sha256":"5718f07d21edd62047391d2515ed5127f79f1f8715dc00377806d203d9d04090","shape":[1,16,21,46,82]}},"t2v":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"edc38a275746d7496fa00e90d2a0ba17e66955d03d1eba92eaf024b732fae80a","dtype":"torch.float32","finite":true,"label":"generated_t2v","numel":1267392,"raw_storage_sha256":"6fedaeb17291e306b62822478503eccc85ae4a921c10aa3ec9de65874bd4d807","shape":[1,16,21,46,82]}}},"initial_noise_artifacts":{"rv2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"8e26e2774c3334500f1e28fff468a12f9610f73dfb627888e689f22068537982","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_rv2v","numel":1267392,"raw_storage_sha256":"2bc2848ee815333fa0b799f191ac39d42be9cfd6ef050b340fe02ce6328c7f91","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"8e26e2774c3334500f1e28fff468a12f9610f73dfb627888e689f22068537982","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081901,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/rv2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"2bc2848ee815333fa0b799f191ac39d42be9cfd6ef050b340fe02ce6328c7f91","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"80f61359e9dab41655add1772843bb06f0c0a150b53201e90c123340076545a3","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"2bc2848ee815333fa0b799f191ac39d42be9cfd6ef050b340fe02ce6328c7f91"},"t2v":{"all_rank_identity":{"all_rank_exact":true,"identity":{"byte_count":5069568,"content_sha256":"8e26e2774c3334500f1e28fff468a12f9610f73dfb627888e689f22068537982","dtype":"torch.float32","finite":true,"label":"official_initial_gaussian_t2v","numel":1267392,"raw_storage_sha256":"2bc2848ee815333fa0b799f191ac39d42be9cfd6ef050b340fe02ce6328c7f91","shape":[1,16,21,46,82]}},"byte_count":5069568,"captured_from_native_sampler":true,"content_sha256":"8e26e2774c3334500f1e28fff468a12f9610f73dfb627888e689f22068537982","coordinate":"bernini_native_target_latent_before_rearrange","dtype":"torch.float32","external_initial_noise_injection":false,"generator_device":"cpu","generator_initial_seed":2026081901,"numel":1267392,"observer_added_device_to_cpu_readback":true,"observer_changed_return_value":false,"observer_only":true,"official_module_global_symbol":"bernini.models.wan_diffusion.randn_tensor","official_randn_tensor_call_count":1,"origin":"observed_return_of_official_module_global_randn_tensor","original_callable_invoked_once_with_unchanged_arguments":true,"original_device":"cuda:0","original_return_tensor_forwarded_by_identity":true,"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/t2v.official-initial-gaussian.safetensors","randn_tensor_call_count":1,"raw_value_sha256":"2bc2848ee815333fa0b799f191ac39d42be9cfd6ef050b340fe02ce6328c7f91","requested_device":"cuda:0","requested_dtype":"torch.float32","roundtrip_raw_value_exact":true,"sampler_noise_replacement":false,"sha256":"4d021febf8e301d05621bcdfa8e3206208560a0ad22191f2a3b180b5cfd87489","shape":[1,16,21,46,82],"source_or_target_derived":false,"stored_device":"cpu","stored_dtype":"torch.float32","tensor_key":"official_initial_gaussian","tensor_value_sha256":"2bc2848ee815333fa0b799f191ac39d42be9cfd6ef050b340fe02ce6328c7f91"}},"input":{"accepted_external_conditions":["source_video","action_prompt"],"action_prompt_utf8_bytes":283,"action_prompt_utf8_sha256":"bd72848533d16119920aa2d3e0b34a8ea82f3fad6c6f39444239e22523d11397","external_first_frame_anchor":false,"external_mask_flow_pose_track_trajectory":false,"external_reference_image_or_video":false,"source_video_path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/preprocessed_sources/fb7725c351ba/source-exact81.mp4","source_video_sha256":"479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a","target_video":false},"interpretation":{"best_arm_selected":false,"purpose":"test_native_identity_conditioned_generation_before_training","quality_claim":false,"training_performed":false},"latent_geometry":{"one_reference_patch_tokens":943,"per_arm_total_visual_tokens":{"r2v":24518,"rv2v":43378,"t2v":19803},"reference_latent_shape":[1,16,1,46,82],"target_patch_tokens":19803,"video_latent_shape":[1,16,21,46,82]},"method":"frozen-bernini-native-identity-generation-canary","method_source_archive_sha256":"fc4a5b08df0da1ba264543753f356d116174f507cdc3b6ba45b55535c7f09619","method_source_revision":"f47f9d3630142d7c20bc6c163a52405aac249a4f","outputs":{"rv2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/rv2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"83881ae8d5619c4ee3c5a65fa0c8ae72a04bee424ef55f3e0910e6fe2379f892","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/rv2v.mp4","sha256":"87307f240a0e050c8f680b4eac960017786cf75a0820c8259b9352b6d72b2697","width":656},"t2v":{"fps":25,"frame_count":81,"height":368,"normalized_clean_latent":{"artifact_role":"native_sampler_proposal","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":true,"origin":"native_sampler_before_vae_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/t2v.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"5fd51078c80f15fff389d0d400d841905760a642ece9c3896faa34fc39f512d0","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":false,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/t2v.mp4","sha256":"bad550261470a3d6befacf71a1d41ee5923dc74764aaa990f4e164fcac9a000c","width":656}},"preprocessing":{"decoded_from_private_byte_snapshot":true,"external_shared_i0":false,"fps":25.0,"frame_count":81,"max_pixels":245760,"original_post_snapshot_sha256":"479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a","original_pre_snapshot_sha256":"479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a","original_stat_identity_stable":true,"reported_fps":25.0,"resize":"torchvision_bicubic_antialias_true","snapshot_sha256":"479ee891773dfffc3299149103ee2ef37f4bb63f213b1c4c7cbd5391ebd4704a","source_derived_bucket_hw":[368,656],"source_input_hw":[720,1280],"spatial_policy":"sqrt_max_pixels_then_floor_each_dimension_to_stride","stride":16,"temporal_policy":"all_integer_frames_0_through_80_no_subsampling"},"production_claim_forbidden":true,"prompt_contract":{"rv2v":{"binding_clause_sha256":"8574f7e8b0f996a84da5e01fc31154dfcd197da2d371800f40f03bbfa0e28d4b","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"415b29b662b33ef06f3e969637acb45a29bd5087fd1820764ed6195e848d89d6","guidance_mode":"rv2v","inference_arm":"rv2v","system_prompt_sha256":"de4b6a32c46d3e9f1172c2801f8f29c18a448f7b1f3c907d91b3f87c4870aca2","tokenizer_fix_mistral_regex":true,"training_task_name":"vr2v"},"t2v":{"binding_clause_sha256":"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855","cleaner":"diffusers.pipelines.wan.pipeline_wan.prompt_clean","full_prompt_sha256":"bca668cb7c69fd3dbcb1ba6823f0196d601fd967f1210b44ca1a6fa468dba78b","guidance_mode":"t2v_apg","inference_arm":"t2v","system_prompt_sha256":"31641ecb2329849be215b334ee550241479e84b3306e148369f6a076b23002e8","tokenizer_fix_mistral_regex":true,"training_task_name":"t2v"}},"receipt_digest":"9572eb83bd3c5e2e5e83bf1fbcd4b90b9ace53a6e1a222c1e1f4f37b2a9b335d","runtime_versions":{"diffusers":"0.38.0","torch":"2.7.1+rocm6.3","torch_hip":"6.3.42131-fa1d09cbd","transformers":"5.5.4"},"sampling":{"rv2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"rv2v","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081901,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4},"t2v":{"custom_sampler_or_scheduler":false,"eta":0.5,"flow_shift":5.0,"guidance_mode":"t2v_apg","momentum":0.0,"norm_threshold":[50.0,50.0],"num_frames":81,"num_inference_steps":40,"omega_img":4.5,"omega_scale":0.8,"omega_txt":4.0,"omega_vid":1.25,"same_seed_and_target_shape_across_arms":true,"seed":2026081901,"single_expert":"transformer_1","target_initialization":"official_gen_wanx22_fresh_gaussian","target_mixed_with_source_latent":false,"ulysses_size":4}},"schema_version":"bernini-native-identity-generation-canary-v1","scientific_claim_authorized":false,"source_condition_artifact":{"artifact_role":"source_video_condition","coordinate":"bernini_normalized_clean_vae_latent","mp4_decode_reencode_used":false,"native_sampler_before_vae_decode":false,"origin":"source_video_vae_encode_before_any_decode","path":"/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/source.normalized-clean-latent.safetensors","roundtrip_byte_exact_fp32":true,"sampler_return_dtype":"torch.float32","sha256":"0f7f845f1edcb7af7bdcf32c7864946527f388070ceb4f31f42415de538e2b08","shape":[1,16,21,46,82],"source_video_vae_encode_before_any_decode":true,"stored_dtype":"torch.float32","tensor_key":"normalized_clean_latent"},"veomni_commit":"f90b3dc6fbb0ce693745223cc7a94064123dbf4d"}
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/bernini_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/bernini_t2v.jpg
new file mode 100644
index 0000000..c586d3c
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/bernini_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/bernini_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/bernini_t2v.mp4
new file mode 100644
index 0000000..45d0cf2
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/bernini_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/launch.json b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/launch.json
new file mode 100644
index 0000000..e1ee5b7
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/launch.json
@@ -0,0 +1,131 @@
+{
+ "case": {
+ "index": 6,
+ "case_id": "fb7725c351ba",
+ "name": "提杠铃 · Barbell lift",
+ "page": null,
+ "seed": 2026081901,
+ "prompt": "A woman in black activewear performs barbell deadlifts with focused determination in a well-equipped home gym. In this continuous shot, The woman in a home gym lifts a barbell from the floor to her waist. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "bd72848533d16119920aa2d3e0b34a8ea82f3fad6c6f39444239e22523d11397",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/t2v.mp4",
+ "bernini_t2v_sha256": "bad550261470a3d6befacf71a1d41ee5923dc74764aaa990f4e164fcac9a000c",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/fb7725c351ba/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081901",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/fb7725c351ba/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "afda3b5eae7e6735b7e51101193b6da532c3845d203e519519e24c1d0488b73b",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:06:59.576739+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081901
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection."
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/omni_t2v.jpg b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/omni_t2v.jpg
new file mode 100644
index 0000000..920c002
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/omni_t2v.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/omni_t2v.mp4 b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/omni_t2v.mp4
new file mode 100644
index 0000000..c763267
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/omni_t2v.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/real_target.jpg b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/real_target.jpg
new file mode 100644
index 0000000..7f8f852
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/real_target.jpg differ
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/real_target.mp4 b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/real_target.mp4
new file mode 100644
index 0000000..964d873
Binary files /dev/null and b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/real_target.mp4 differ
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/real_target_receipt.json b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/real_target_receipt.json
new file mode 100644
index 0000000..1bdf92c
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/real_target_receipt.json
@@ -0,0 +1,58 @@
+{
+ "case_id": "fb7725c351ba",
+ "collected_utc": "2026-09-23T17:35:12.301202+00:00",
+ "purpose": "review_only",
+ "used_for_inference": false,
+ "original": {
+ "path": "/vast/users/guangyi.chen/dataset/MEV/MEV/videos/11ea4341-888a-41c7-9482-49046d1e113b-00.00.01.984-00.00.04.353-seg2.mp4",
+ "sha256": "5b49f9be6549c6fef6cc803eb278b83e9cbaa7f1ef8de299d24433ba79dd1b3c"
+ },
+ "original_probe": {
+ "width": 1280,
+ "height": 720,
+ "avg_frame_rate": "25/1",
+ "duration": "2.440000",
+ "nb_read_frames": "61"
+ },
+ "display": {
+ "path": "results/fb7725c351ba/real_target.mp4",
+ "sha256": "4c783a0078c94ec89c48be96a9ccf4cdf736ac5f7593164cf86b79bdaeb0f88b",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ },
+ "ffmpeg_command": [
+ "ffmpeg",
+ "-hide_banner",
+ "-loglevel",
+ "error",
+ "-y",
+ "-i",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/real_targets_original/fb7725c351ba.mp4",
+ "-map",
+ "0:v:0",
+ "-an",
+ "-vf",
+ "setpts=(PTS-STARTPTS)*1.32786885245902,fps=25,tpad=stop_mode=clone:stop_duration=3.24,trim=end_frame=81,setpts=N/(25*TB),scale=960:540:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=960:540:(ow-iw)/2:(oh-ih)/2,setsar=1",
+ "-frames:v",
+ "81",
+ "-c:v",
+ "libx264",
+ "-crf",
+ "18",
+ "-preset",
+ "fast",
+ "-threads",
+ "2",
+ "-pix_fmt",
+ "yuv420p",
+ "-movflags",
+ "+faststart",
+ "/Users/siriuschu/ML/VideoEditing/VideoEdit/artifacts/progress2_omni_t2v12_20260923/results/fb7725c351ba/real_target.mp4"
+ ],
+ "full_decode_passed": true
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/receipt.json b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/receipt.json
new file mode 100644
index 0000000..2cbd934
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/receipt.json
@@ -0,0 +1,175 @@
+{
+ "status": "complete",
+ "case": {
+ "index": 6,
+ "case_id": "fb7725c351ba",
+ "name": "提杠铃 · Barbell lift",
+ "page": null,
+ "seed": 2026081901,
+ "prompt": "A woman in black activewear performs barbell deadlifts with focused determination in a well-equipped home gym. In this continuous shot, The woman in a home gym lifts a barbell from the floor to her waist. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "prompt_sha256": "bd72848533d16119920aa2d3e0b34a8ea82f3fad6c6f39444239e22523d11397",
+ "bernini_t2v": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/t2v.mp4",
+ "bernini_t2v_sha256": "bad550261470a3d6befacf71a1d41ee5923dc74764aaa990f4e164fcac9a000c",
+ "bernini_receipt": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/mev_action_anchor_target_gap16_20260819_v1/generation/fb7725c351ba/receipt.json",
+ "comparison_batch": "progress2_omni6_extra_20260923"
+ },
+ "command": [
+ "/vast/users/guangyi.chen/anaconda3/envs/vace/bin/python3.12",
+ "-m",
+ "torch.distributed.run",
+ "--standalone",
+ "--nproc-per-node",
+ "1",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/fb7725c351ba/attempt-001/omni_runtime/tools/inference/generate_omni_v2v_1_3B.py",
+ "--task",
+ "t2v-1.3B",
+ "--size",
+ "832*480",
+ "--frame_num",
+ "81",
+ "--sample_fps",
+ "25",
+ "--sample_shift",
+ "5",
+ "--sample_solver",
+ "unipc",
+ "--sample_steps",
+ "40",
+ "--sample_guide_scale",
+ "3",
+ "--base_seed",
+ "2026081901",
+ "--classifier_free_ratio",
+ "0",
+ "--ckpt_dir",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "--new_checkpoint",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e/transformer/pytorch_model.pt",
+ "--use_usp",
+ "false",
+ "--sp_size",
+ "1",
+ "--t5_fsdp",
+ "false",
+ "--dit_fsdp",
+ "false",
+ "--max_context_len",
+ "9216",
+ "--qwen3vl_dtype",
+ "bf16",
+ "--qwen3vl_device_map",
+ "balanced_low_0",
+ "--qwen3vl_temperature",
+ "0",
+ "--qwen3vl_model_path",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "--prompt_file",
+ "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/work/fb7725c351ba/attempt-001/text_only_input.jsonl"
+ ],
+ "model": {
+ "family": "Fudan-FUXI/OmniVideo2-1.3B",
+ "checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_shared8_20260805_10932d4/models/OmniVideo2-1.3B-composite-d994028e",
+ "transformer_sha256": "f269fe8c6b35993bbb4ea340c535ee9893928ea215fb8c4be3d5e9f122d844d6",
+ "special_tokens_sha256": "72129ce9ade25aa0fbf738c005d3dc090c1b6c45918580e1c683b6ecef726ad4",
+ "umt5_checkpoint_sha256": "7cace0da2b446bbbbc57d031ab6cf163a3d59b366da94e5afe36745b746fd81d",
+ "vae_checkpoint_sha256": "38071ab59bd94681c686fa51d75a1968f64e470262043be31f7a094e442fd981",
+ "umt5_plus_tokenizer_prior_identity_manifest_sha256": "99bec3d4e1a5c50694eb95d3fbec880d32bac8c0536c3dea8080b459f90e32cd",
+ "qwen_checkpoint_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/marp_omnivideo2_maskfree/models/Qwen3-VL-30B-A3B-Instruct",
+ "qwen_prior_identity_manifest_sha256": "5a0c83274293904863d5aed304385e906afa02dc04d36e9b34ac8383775b78be",
+ "qwen_prior_identity_file_count": 25,
+ "qwen_prior_identity_total_bytes": 62153207155,
+ "adaptation": "official_base_no_project_MARP_adapter",
+ "official_revision": "adcee0a4a5b439ad3615f825298221b21177d4e3",
+ "official_archive_sha256": "2095a1b0449215fcb8151aac76584b6b2fb4a99541e30e0bb986901a81007ffd",
+ "official_unified_model_sha256": "739addc6b46c62fb389cf037736d52837d8e0fc911d84e3ad7db05e094ee7328",
+ "official_generate_entry_sha256": "f7cb67288d9a860a92fa0a548b139d6a15194b1a7db837b04889dc8d75b36be6",
+ "patched_unified_model_sha256": "7d1dee31b4cb2632cce7ee722e3ce2f43fbbf75c6f93203aaccf5b568efbba0d",
+ "patched_generate_entry_sha256": "d46a0b3863944689c1566722b299745f531c6baacafa67f9902a4101e4517973",
+ "runtime_patch_contract": [
+ "T2V entrypoint adaptation for single-model 1.3B",
+ "strict-equivalent checkpoint load",
+ "reject all input-media reads and source VAE encoding",
+ "reject visual tokens and context truncation"
+ ],
+ "official_t2v_entry_sha256": "f3e3ad47d37e436a6a633da07e5156dcd9238d635f94f4470d289067942d0284",
+ "official_vlm_helper_sha256": "8aa7e97603094dfeeb5e9b02bfe294d7dce3fad780eed1e43cc044efa56d295f",
+ "transplanted_t2v_conditioning_block_sha256": "94d0f178304d21781651cb26386d8ea8f191f872f9e8424912461cb6bfea4279",
+ "entrypoint_adaptation": "Official T2V Qwen/T5 conditioning block transplanted unchanged into official 1.3B single-checkpoint entry; no model/sampler changes."
+ },
+ "manifest_sha256": "e8ad3586da335f84e80987c146fb54347127cd9e0d57b89c1ded65d12c65ded5",
+ "driver_sha256": "e79d83272c2286c25f01674c11270081ed21fbafe4844bdc6b6c16a2c1ddf453",
+ "runtime_adapter_sha256": "fff4529cc4e52dde617c2cb8b056cdd8772985520bf90407b27baab0f9be0770",
+ "prompt_file_sha256": "afda3b5eae7e6735b7e51101193b6da532c3845d203e519519e24c1d0488b73b",
+ "hostname": "auh7-1b-gpu-287",
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "started_utc": "2026-09-23T08:06:59.576739+00:00",
+ "inference_inputs": [
+ "text_prompt"
+ ],
+ "source_video_used": false,
+ "real_target_used": false,
+ "reference_media_used": false,
+ "sampler": {
+ "name": "unipc",
+ "steps": 40,
+ "cfg": 3,
+ "flow_shift": 5,
+ "seed": 2026081901
+ },
+ "geometry": {
+ "width": 832,
+ "height": 480,
+ "frames": 81,
+ "fps": 25
+ },
+ "text_contract": "Exact Bernini T2V body caption; native Omni Qwen text-only expansion and features, then T5. No Bernini task-prefix injection.",
+ "conditioning": {
+ "input_prompt": "A woman in black activewear performs barbell deadlifts with focused determination in a well-equipped home gym. In this continuous shot, The woman in a home gym lifts a barbell from the floor to her waist. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A woman in sleek black activewear performs a series of controlled barbell deadlifts in a well-equipped home gym, her movements precise and deliberate. The camera holds a steady, medium-wide shot, capturing her full body as she stands with feet shoulder-width apart, gripping the barbell with a firm overhand hold. With focused determination etched on her face, she bends at the hips and knees, lowering the barbell smoothly from the floor to her waist, maintaining a straight back and engaged core throughout. The gym is bathed in soft, even lighting that highlights the clean, organized space—treadmills, weight racks, and resistance bands visible in the background. The scene remains consistent, with no changes in framing, camera angle, or lighting, emphasizing the continuity and rhythm of her exercise. Her breath is steady, her muscles taut, as she repeats the motion with unwavering form, the metallic clink of the barbell echoing faintly in the quiet room. The atmosphere is one of quiet intensity, underscoring her commitment and strength.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 65,
+ 2048
+ ],
+ "t5_context_shape": [
+ 243,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 322
+ ],
+ "rank_zero_log_sha256": "329d3cbfb8739a2bd3f2683b42a4efd725f0e02a935aa0908ae813b354bd02a9"
+ },
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/fb7725c351ba/omni_t2v.mp4",
+ "sha256": "032ff8669d17b5b23b66c8f92b0119b23484b8011eed7238f7509b3d8c77b6f9",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "full_decode_passed": true,
+ "elapsed_seconds": 773.34,
+ "completed_utc": "2026-09-23T08:19:13.031682+00:00"
+}
diff --git a/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/t2v_conditioning.json b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/t2v_conditioning.json
new file mode 100644
index 0000000..3909eb1
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/results/fb7725c351ba/t2v_conditioning.json
@@ -0,0 +1,19 @@
+{
+ "input_prompt": "A woman in black activewear performs barbell deadlifts with focused determination in a well-equipped home gym. In this continuous shot, The woman in a home gym lifts a barbell from the floor to her waist. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A woman in sleek black activewear performs a series of controlled barbell deadlifts in a well-equipped home gym, her movements precise and deliberate. The camera holds a steady, medium-wide shot, capturing her full body as she stands with feet shoulder-width apart, gripping the barbell with a firm overhand hold. With focused determination etched on her face, she bends at the hips and knees, lowering the barbell smoothly from the floor to her waist, maintaining a straight back and engaged core throughout. The gym is bathed in soft, even lighting that highlights the clean, organized space\u2014treadmills, weight racks, and resistance bands visible in the background. The scene remains consistent, with no changes in framing, camera angle, or lighting, emphasizing the continuity and rhythm of her exercise. Her breath is steady, her muscles taut, as she repeats the motion with unwavering form, the metallic clink of the barbell echoing faintly in the quiet room. The atmosphere is one of quiet intensity, underscoring her commitment and strength.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 65,
+ 2048
+ ],
+ "t5_context_shape": [
+ 243,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+}
\ No newline at end of file
diff --git a/projects/video-action-editing/omni-t2v12/summary.json b/projects/video-action-editing/omni-t2v12/summary.json
new file mode 100644
index 0000000..0ce66ab
--- /dev/null
+++ b/projects/video-action-editing/omni-t2v12/summary.json
@@ -0,0 +1,782 @@
+[
+ {
+ "case_id": "fb7725c351ba",
+ "name": "提杠铃 · Barbell lift",
+ "seed": 2026081901,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/fb7725c351ba/omni_t2v.mp4",
+ "sha256": "032ff8669d17b5b23b66c8f92b0119b23484b8011eed7238f7509b3d8c77b6f9",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "hostname": "auh7-1b-gpu-287",
+ "elapsed_seconds": 773.34,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 322
+ ],
+ "rank_zero_log_sha256": "329d3cbfb8739a2bd3f2683b42a4efd725f0e02a935aa0908ae813b354bd02a9"
+ },
+ "conditioning": {
+ "input_prompt": "A woman in black activewear performs barbell deadlifts with focused determination in a well-equipped home gym. In this continuous shot, The woman in a home gym lifts a barbell from the floor to her waist. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A woman in sleek black activewear performs a series of controlled barbell deadlifts in a well-equipped home gym, her movements precise and deliberate. The camera holds a steady, medium-wide shot, capturing her full body as she stands with feet shoulder-width apart, gripping the barbell with a firm overhand hold. With focused determination etched on her face, she bends at the hips and knees, lowering the barbell smoothly from the floor to her waist, maintaining a straight back and engaged core throughout. The gym is bathed in soft, even lighting that highlights the clean, organized space—treadmills, weight racks, and resistance bands visible in the background. The scene remains consistent, with no changes in framing, camera angle, or lighting, emphasizing the continuity and rhythm of her exercise. Her breath is steady, her muscles taut, as she repeats the motion with unwavering form, the metallic clink of the barbell echoing faintly in the quiet room. The atmosphere is one of quiet intensity, underscoring her commitment and strength.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 65,
+ 2048
+ ],
+ "t5_context_shape": [
+ 243,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/fb7725c351ba/real_target.mp4",
+ "sha256": "4c783a0078c94ec89c48be96a9ccf4cdf736ac5f7593164cf86b79bdaeb0f88b",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "3fadbe0c2961",
+ "name": "狗转头 · Dog head turn",
+ "seed": 2026081902,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/3fadbe0c2961/omni_t2v.mp4",
+ "sha256": "6b28c93fda92e5fa29cfe7e1954ee29dd4c6e1d594b657d7a9cf7df195b67e56",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "hostname": "auh7-1b-gpu-287",
+ "elapsed_seconds": 555.7,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 341
+ ],
+ "rank_zero_log_sha256": "2a6eb521a55ce5cdc05902458c5a2b863cb3532531dd394b6fc5278eac4aa7db"
+ },
+ "conditioning": {
+ "input_prompt": "A happy Pembroke Welsh Corgi with a reddish-brown and white coat sits in a sunny green field, panting with its tongue out in a static medium shot. In this continuous shot, The dog turns its head further to the right while sitting in the grass. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A happy Pembroke Welsh Corgi with a rich reddish-brown and white coat sits calmly in a sun-drenched green field, its short legs tucked beneath its sturdy body. The warm sunlight bathes the scene, casting soft highlights across the dog’s smooth fur and illuminating the vibrant blades of grass around it. The corgi pants gently, its pink tongue lolling out in a relaxed, joyful expression, conveying a sense of contentment and ease. The camera holds a steady, static medium shot, framing the dog from the mid-chest up, allowing full view of its expressive face and perked ears. As the moment unfolds, the dog slowly turns its head further to the right, its eyes briefly scanning the surroundings with quiet curiosity, while its body remains still. The lush green field stretches out behind it, dotted with wildflowers and bathed in golden afternoon light. The scene remains consistent throughout—same subject, same environment, same lighting, same framing—capturing a peaceful, serene moment in nature, with the corgi as the calm, cheerful centerpiece.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 77,
+ 2048
+ ],
+ "t5_context_shape": [
+ 250,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/3fadbe0c2961/real_target.mp4",
+ "sha256": "2d97f922dbbaab42d8fb0ff8419410dc7275ddf07d301b1113871e2d651c1e4f",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "7077cfb160cf",
+ "name": "长椅上写字 · Write in notebook",
+ "seed": 2026081904,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/7077cfb160cf/omni_t2v.mp4",
+ "sha256": "061290c54920d2b552b84b511873f10ead41e881d0f091fc11d0323e97e96427",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "hostname": "auh7-1b-gpu-287",
+ "elapsed_seconds": 540.8,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 320
+ ],
+ "rank_zero_log_sha256": "2974e23fa644b0fa38f0a8091e8bf9fd9ef15ef20ffdb210b14d8a3d1aa75546"
+ },
+ "conditioning": {
+ "input_prompt": "A young woman with shoulder-length light brown hair sits on a wooden bench in a sunny park or campus setting, writing in a notebook while dressed in a pink shirt and blue jeans. In this continuous shot, The young woman sitting on a park bench thoughtfully writes in her notebook with a pen. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young woman with shoulder-length light brown hair sits calmly on a weathered wooden bench in a sun-drenched park or campus setting, the golden light of midday filtering through the leaves of nearby trees. She wears a soft pink button-up shirt and fitted blue jeans, her posture relaxed yet focused as she leans slightly forward, pen in hand, writing thoughtfully in a small notebook resting on her lap. Her brow is gently furrowed in concentration, occasionally pausing to glance up and around, as if gathering inspiration from the peaceful surroundings. The camera holds steady in a medium shot, capturing her from the waist up, with the background softly blurred to emphasize her presence—dappled sunlight dances across the grass and pavement, and the faint rustle of leaves and distant chatter of people passing by add a gentle soundtrack to the scene. The composition remains unchanged throughout, maintaining consistent framing, lighting, and perspective, creating a serene, contemplative moment frozen in time.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 80,
+ 2048
+ ],
+ "t5_context_shape": [
+ 226,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/7077cfb160cf/real_target.mp4",
+ "sha256": "03a7b92dc3800e8403ce9f7f4ce85b9668a85a82e110e0356e614d4626496bc3",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "7e488525099e",
+ "name": "转身倚车 · Turn and lean on car",
+ "seed": 2026081906,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/7e488525099e/omni_t2v.mp4",
+ "sha256": "bf577e972362491fe79f36079915ac95bd4110ff4a2278184e1c284260ad64e2",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "hostname": "auh7-1b-gpu-287",
+ "elapsed_seconds": 772.82,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 346
+ ],
+ "rank_zero_log_sha256": "2cb8628e967b63c4350488a3109cc8bf44c4e79d0bf6ed4828dc3a9486f3b76e"
+ },
+ "conditioning": {
+ "input_prompt": "A confident woman in red pants and a black leather jacket poses in front of a black Chevrolet Camaro in a moody, red-lit underground parking garage. In this continuous shot, The woman in a black leather jacket and red pants turns and leans back against the hood of the black car. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A confident woman stands in front of a sleek black Chevrolet Camaro in a moody, dimly lit underground parking garage, bathed in a deep, ambient red glow that casts dramatic shadows across the concrete walls and floor. She wears a form-fitting black leather jacket and vibrant red pants that contrast sharply with the dark surroundings, drawing the eye to her poised stance. The camera holds steady in a continuous, smooth shot, capturing her as she turns with deliberate grace, her movement fluid and intentional. She leans back against the curved hood of the car, her body angled slightly, one hand resting lightly on the glossy surface while the other hangs relaxed at her side. The red lighting pulses subtly, reflecting off the car’s polished paint and the woman’s jacket, enhancing the cinematic atmosphere. The scene remains tightly framed, with the Camaro’s aggressive lines and the woman’s commanding presence forming a powerful visual composition. The lighting, framing, and camera position remain consistent throughout, emphasizing the stillness and intensity of the moment, as if frozen in time.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 80,
+ 2048
+ ],
+ "t5_context_shape": [
+ 252,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/7e488525099e/real_target.mp4",
+ "sha256": "30c8909ffd3c044c78714102c8d24dc38cf3547c123f0be2deb43390a5732ec3",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "d5221a1b69d5",
+ "name": "闭眼尝汤 · Taste from ladle",
+ "seed": 2026081907,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/d5221a1b69d5/omni_t2v.mp4",
+ "sha256": "77cbffd0f6650d641829a18b2f5dc7244c69ee0d430ca1f5e2a5efaa44e984e2",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "hostname": "auh7-1b-gpu-287",
+ "elapsed_seconds": 552.79,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 358
+ ],
+ "rank_zero_log_sha256": "553659a0edaf6f48af82c44481a5de73ec1c71aac3c3a88556e15fb616195053"
+ },
+ "conditioning": {
+ "input_prompt": "A young woman in a patterned apron stands in a modern kitchen, tasting food from a ladle with a satisfied expression. In this continuous shot, The woman in an apron closes her eyes as she tastes from a metal ladle. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young woman stands in a bright, modern kitchen, her figure centered in a steady, continuous shot. She wears a patterned apron over casual clothing, the fabric soft and slightly rumpled from use. The kitchen is sleek and well-lit, with natural light streaming through a large window, casting a warm glow across stainless steel appliances and clean white countertops. In her hand, she holds a metal ladle, its surface gleaming under the light. She lifts it slowly to her lips, her eyes closing gently as she tastes the food. Her expression softens into quiet satisfaction—lips parting slightly, a subtle smile forming as the flavor registers. The camera remains fixed, capturing the moment with unwavering focus, preserving the stillness of the scene. The lighting remains consistent, highlighting the contours of her face and the texture of the apron. The background remains unchanged—neatly arranged utensils, a pot simmering on the stove, a bowl of fresh herbs on the counter—all contributing to the calm, domestic atmosphere. The shot lingers, emphasizing the sensory experience, the quiet joy of a meal well-prepared, and the simple pleasure of tasting one’s own creation.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 70,
+ 2048
+ ],
+ "t5_context_shape": [
+ 274,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/d5221a1b69d5/real_target.mp4",
+ "sha256": "3b61c8123a193d9feab392a7780ea5b87b91099a2c41ea069875febc81ea9f4e",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "40712e1341dc",
+ "name": "放下手机后翻身睡觉 · Put phone away and sleep",
+ "seed": 2026081911,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/40712e1341dc/omni_t2v.mp4",
+ "sha256": "5ed0844a80af94da135f1290f1307c584e6c5ce29418f8a90f3d940405b2ca11",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213688",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "hostname": "auh7-1b-gpu-287",
+ "elapsed_seconds": 539.42,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 333
+ ],
+ "rank_zero_log_sha256": "762bda731c46287b756b96674e0f3034c5c36417e76e21e33835ff39d69d3781"
+ },
+ "conditioning": {
+ "input_prompt": "A young man with dyed-blond hair lies in bed in a dark room, illuminated by his phone's blue light, wearing a pink tank top and a nose ring. In this continuous shot, The man in bed puts his smartphone on the nightstand and turns over to go to sleep. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young man with dyed-blond hair lies on his back in a dimly lit bedroom, the only source of light emanating from the cool blue glow of his smartphone screen. He wears a soft pink tank top, the fabric slightly rumpled from lying still, and a small nose ring glints faintly in the low light as he shifts his head. The room is dark, shadows pooling in the corners, with only the faint outline of a nightstand and a pillow visible. The camera holds a steady, fixed position, capturing the scene in a continuous, unbroken shot. The man’s fingers move slowly across the phone’s screen, scrolling or typing with quiet focus. After a moment, he lowers the device, places it gently on the nightstand beside him, and turns onto his side, pulling the blanket slightly higher. His breathing slows, and he settles into sleep, the blue light fading as the phone powers down. The scene remains unchanged in framing, lighting, and composition, maintaining a calm, intimate stillness throughout.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 80,
+ 2048
+ ],
+ "t5_context_shape": [
+ 239,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/40712e1341dc/real_target.mp4",
+ "sha256": "4424517beb4e790280cbf782fa1a2d6589b0443d14b9adfb83922d82d919c93e",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "10af6c651dbc",
+ "name": "Jellyfish contraction",
+ "seed": 2026081910,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/10af6c651dbc/omni_t2v.mp4",
+ "sha256": "2380b955b966ff8643bf20f80996659d6c9b2f986743da29a3a2136941ed4437",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "hostname": "auh7-1b-gpu-253",
+ "elapsed_seconds": 779.17,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 334
+ ],
+ "rank_zero_log_sha256": "1253314f59ddf9f03be7e95cfc2dd23bac0a7d266cbca65d56c57593ab4e1d71"
+ },
+ "conditioning": {
+ "input_prompt": "A small, white-spotted jellyfish drifts and pulsates gracefully in serene, deep blue water. In this continuous shot, The jellyfish pulsates, its bell contracting to reveal a ring-like shape in the blue water. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A small, white-spotted jellyfish drifts effortlessly through serene, deep blue water, its translucent bell gently pulsating in rhythmic motion. The delicate, rhythmic contractions of its body propel it forward in a slow, graceful dance, each pulse creating a subtle ripple in the surrounding water. As the bell contracts, it reveals a distinct ring-like shape, a luminous outline that glows faintly against the rich, uniform blue backdrop. The white spots on its body shimmer faintly with the play of light filtering down from above, adding a soft, ethereal contrast to the deep oceanic hue. The scene remains calm and undisturbed, with no sudden movements or changes in the environment. The camera holds steady, maintaining a consistent framing that centers the jellyfish in the frame, capturing its elegant motion in a continuous, uninterrupted shot. The lighting remains soft and diffused, enhancing the tranquil, otherworldly atmosphere of the deep sea. The jellyfish continues its slow, mesmerizing journey, its form and movement unchanged, creating a peaceful and hypnotic visual experience.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 67,
+ 2048
+ ],
+ "t5_context_shape": [
+ 253,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/10af6c651dbc/real_target.mp4",
+ "sha256": "c848feb636d4f40415224a26f79cc88659a34d63506b1ea2d95fecb6a7a01542",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "7fa338474c75",
+ "name": "Playful selfie",
+ "seed": 2026081900,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/7fa338474c75/omni_t2v.mp4",
+ "sha256": "b4e27999f2701c65ac62157ae6f49539ce092acc1dba0f652532fe793b037fc3",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "hostname": "auh7-1b-gpu-253",
+ "elapsed_seconds": 548.75,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 348
+ ],
+ "rank_zero_log_sha256": "bfa9dc5a39eb59949e0310fbc226255b473cd5ca45931ddeeb75709e0c0c89bb"
+ },
+ "conditioning": {
+ "input_prompt": "A cheerful woman in a red bikini top and straw sun hat takes a playful selfie against a green screen background in a static medium shot. In this continuous shot, The woman in a red bikini top and straw hat takes a playful selfie with her phone, sticking her tongue out and winking. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A cheerful woman stands in a static medium shot against a bright green screen background, her figure centered and clearly defined. She wears a vibrant red bikini top and a wide-brimmed straw sun hat that casts a soft shadow over her face, enhancing the playful, summery vibe. Her expression is lively and animated as she holds her smartphone at arm’s length, capturing a selfie with a mischievous grin. She sticks her tongue out slightly and winks one eye, adding a flirtatious, carefree energy to the moment. Her dark hair peeks out from under the hat, gently swaying as she shifts her weight slightly, maintaining a relaxed yet engaging posture. The lighting is even and bright, highlighting her features and the vivid colors of her outfit without casting harsh shadows. The camera remains completely still throughout, keeping the framing consistent and focused on her upper body and face. The green screen behind her is uniform and uncluttered, providing a clean, neutral backdrop that emphasizes her presence. The scene exudes a lighthearted, fun atmosphere, capturing a spontaneous and joyful moment in a single, continuous take.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 79,
+ 2048
+ ],
+ "t5_context_shape": [
+ 255,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/7fa338474c75/real_target.mp4",
+ "sha256": "8427b71fda1c01fe061b9943820e618c9f887c03a19f79f9191e4ee879b5a25a",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "81533c9e56ec",
+ "name": "Scarf spin",
+ "seed": 2026081915,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/81533c9e56ec/omni_t2v.mp4",
+ "sha256": "11a91b035dac93b7875974b44469c54743c226ea892402e0d742dc90a0e49ca4",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "0,1,2,3"
+ },
+ "hostname": "auh7-1b-gpu-253",
+ "elapsed_seconds": 543.4,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 355
+ ],
+ "rank_zero_log_sha256": "b6de7b091cb33a2ef71fff25d838ebe3334d7999fd50c1abde4bbe69a0ff0e80"
+ },
+ "conditioning": {
+ "input_prompt": "A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house. In this continuous shot, The woman wearing a green skirt and black shirt gracefully spins in a circle while holding a scarf. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A woman in a vintage-style dress dances gracefully by a flower-lined fence in front of a rustic wooden house, her movements fluid and deliberate under the warm, golden light of late afternoon. She wears a flowing green skirt that swirls with each turn, paired with a fitted black shirt that contrasts elegantly with the soft hues of the surrounding garden. In her hands, she holds a long, sheer scarf that billows like a ribbon in the breeze, catching the sunlight as she spins in a slow, continuous circle. The camera remains steady, capturing her in a wide, stable frame that emphasizes the harmony between her motion and the tranquil setting. The fence, draped in climbing roses and wildflowers, frames the scene with natural beauty, while the weathered wooden house in the background adds a sense of timeless charm. The lighting is soft and diffused, casting gentle shadows and enhancing the nostalgic atmosphere. Her expression is serene, eyes closed at times, as if lost in the rhythm of the moment. The entire scene unfolds in a single, unbroken shot, with no cuts or shifts in perspective, allowing the viewer to fully immerse in the grace and stillness of the moment.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 67,
+ 2048
+ ],
+ "t5_context_shape": [
+ 274,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/81533c9e56ec/real_target.mp4",
+ "sha256": "5edb351b04edd07ccb51fc57043298d62a9abcb959d6f86bd5912ef5d654d6d3",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "d977ce147d80",
+ "name": "Rub eyes then cover face",
+ "seed": 2026081905,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/d977ce147d80/omni_t2v.mp4",
+ "sha256": "511b0763dc6b87bd4ecb80dae8ba4b95442b8c424f861e2587297e9841316edd",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "hostname": "auh7-1b-gpu-253",
+ "elapsed_seconds": 778.41,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 399
+ ],
+ "rank_zero_log_sha256": "553e3d91be0dff92c44c637859f9a1428b84512cd4ae3cb64beb94aafe45df43"
+ },
+ "conditioning": {
+ "input_prompt": "A young man with dark skin, dreadlocks, and a gold septum ring lies on a dark couch in a dimly lit room, showing signs of distress. In this continuous shot, The man lying on the couch rubs his tired eyes and covers his face with both hands. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young man with dark skin and long dreadlocks lies sprawled across a dark, plush couch in a dimly lit room, the low ambient light casting deep shadows that pool around him. A gold septum ring glints faintly as he shifts slightly, his body tense with quiet unease. The room is still and intimate, the atmosphere heavy with a sense of emotional weight. He lifts his hands slowly, rubbing his tired eyes with the heels of his palms, his fingers pressing into the sockets as if trying to push back exhaustion or pain. Then, with a slow, deliberate motion, he covers his face entirely, pressing both hands against his forehead and cheeks, his breath shallow and uneven. The camera holds steady, capturing the moment in a continuous, unbroken shot—no movement, no cut—allowing the viewer to sit with the man’s distress. The framing remains tight and consistent, focusing on his upper body and face, emphasizing the subtle tremors in his hands and the way his shoulders rise and fall with each labored breath. The lighting remains unchanged, soft and moody, highlighting the contours of his face and the texture of his hair, while the surrounding darkness absorbs all distractions, drawing attention solely to his internal struggle. The scene is quiet, almost still, yet charged with unspoken emotion.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 78,
+ 2048
+ ],
+ "t5_context_shape": [
+ 307,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/d977ce147d80/real_target.mp4",
+ "sha256": "9a2ad5f5fc5fca2da6532ecddbf6d5879e9c09cfcb57b952b8b28758d628cb1e",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "840b214afead",
+ "name": "Place bottle on treadmill",
+ "seed": 2026081913,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/840b214afead/omni_t2v.mp4",
+ "sha256": "f8c6c74bac6feaeb53623326afaf4ba7a839d47b75d8981bd400429fd9cd5cd1",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "hostname": "auh7-1b-gpu-253",
+ "elapsed_seconds": 555.63,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 385
+ ],
+ "rank_zero_log_sha256": "d865c4c2b75642465de5471b399e7825250690c84f63f207fafa7208dc82ca6d"
+ },
+ "conditioning": {
+ "input_prompt": "A young woman with long reddish-blonde hair, wearing a mint green athletic outfit, stands on a treadmill in a modern gym with blue and white striped pillars, holding a water bottle. In this continuous shot, The woman turns her head to the left and places her water bottle onto the treadmill. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A young woman with long, sun-kissed reddish-blonde hair stands centered on a treadmill in a sleek, modern gym, her posture relaxed yet alert. She wears a form-fitting mint green athletic outfit—a moisture-wicking tank top and matching leggings—that contrasts softly against the cool-toned environment. The gym is illuminated by bright, even lighting that highlights the clean lines of the space, with blue and white striped pillars rising in the background, creating a rhythmic visual pattern that adds depth to the scene. The camera holds a steady, medium-wide shot, capturing her from the waist up, with the treadmill’s control panel and running belt clearly visible. In a smooth, continuous motion, she turns her head slightly to the left, her gaze briefly shifting toward the side of the treadmill, then lowers her right hand to place a clear water bottle with a blue cap onto the treadmill’s console. The bottle rests securely, and she remains in place, her expression calm and focused, as the scene continues without interruption. The lighting remains consistent, the framing stable, and the background elements—pillars, mirrors, and distant gym equipment—remain unchanged, emphasizing the quiet, deliberate nature of the moment.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 82,
+ 2048
+ ],
+ "t5_context_shape": [
+ 289,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/840b214afead/real_target.mp4",
+ "sha256": "989454c22b5f6db63ebd8667c943bb2622cfc2c25990661805dda3e1f9fde52b",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ },
+ {
+ "case_id": "d908552fb691",
+ "name": "Bow to clap",
+ "seed": 2026081903,
+ "output": {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/progress2_omni_t2v12_20260923_v1/results/d908552fb691/omni_t2v.mp4",
+ "sha256": "97ca0d377f030e084df41b86b49ad784496e336a90223c1acd81eb886f8366f6",
+ "codec": "h264",
+ "width": 832,
+ "height": 480,
+ "frame_count": 81,
+ "fps": 25.0,
+ "duration_seconds": 3.24
+ },
+ "job": {
+ "SLURM_JOB_ID": "213687",
+ "SLURM_ARRAY_JOB_ID": null,
+ "SLURM_ARRAY_TASK_ID": null,
+ "SLURM_JOB_GPUS": "0,1,2,3,4,5,6,7",
+ "ROCR_VISIBLE_DEVICES": "4,5,6,7"
+ },
+ "hostname": "auh7-1b-gpu-253",
+ "elapsed_seconds": 533.44,
+ "runtime_evidence": {
+ "strict_equivalent_checkpoint_load": true,
+ "context_truncation": false,
+ "visual_condition_tokens": 0,
+ "context_log_records": 80,
+ "context_total_tokens": [
+ 144,
+ 335
+ ],
+ "rank_zero_log_sha256": "4f029607a0c836fa49e9a318191fdfebd6a1aff88df0fe004ff5b9ba44aafa00"
+ },
+ "conditioning": {
+ "input_prompt": "A woman in a white taekwondo uniform with a black collar and Adidas logo stands in profile against a black background under dramatic side lighting, captured in a static medium shot. In this continuous shot, The woman in a taekwondo uniform stands in profile and claps her hands together in front of her. The same subject identity, scene, lighting, framing, and camera remain stable.",
+ "expanded_caption": "A woman in a crisp white taekwondo uniform stands in profile against a deep black background, her form sharply defined by dramatic side lighting that casts a strong, sculpted shadow along her silhouette. The uniform features a black collar and a subtle Adidas logo on the chest, adding a touch of brand detail to the athletic attire. She is captured in a static medium shot, framed with precision to highlight her poised stance and the clean lines of her posture. The lighting emphasizes the contours of her body, creating a high-contrast, cinematic effect that enhances the intensity of the moment. In a smooth, continuous motion, she brings her hands together in front of her chest, clapping them once with a deliberate, controlled gesture. Her movements are precise and measured, reflecting discipline and focus. The scene remains unchanged throughout—no shifts in camera angle, lighting, or background—maintaining a consistent, composed atmosphere that centers entirely on her presence and action. The stillness of the frame contrasts with the subtle energy of her motion, creating a powerful visual statement of control and readiness.",
+ "source_video_path": null,
+ "visual_embedding": null,
+ "source_video_caption": "",
+ "qwen_features_origin": "text_only",
+ "qwen_feature_shape": [
+ 1,
+ 85,
+ 2048
+ ],
+ "t5_context_shape": [
+ 236,
+ 4096
+ ],
+ "media_read_attempts": 0,
+ "source_vae_encode_attempts": 0
+ },
+ "real_target": {
+ "path": "results/d908552fb691/real_target.mp4",
+ "sha256": "ff161cba09e00e281c6407a3e96e7e359f9900a4d0af4dc6a0d042931aeadcf6",
+ "frames": 81,
+ "fps": 25,
+ "duration": 3.24,
+ "width": 960,
+ "height": 540,
+ "temporal_normalization": "full clip rescaled to 3.24 seconds",
+ "spatial_normalization": "fit and letterbox; no crop"
+ }
+ }
+]
diff --git a/projects/video-action-editing/paper/action_support_recovery.pdf b/projects/video-action-editing/paper/action_support_recovery.pdf
new file mode 100644
index 0000000..fd74c2a
Binary files /dev/null and b/projects/video-action-editing/paper/action_support_recovery.pdf differ
diff --git a/projects/video-action-editing/paper/action_support_recovery.svg b/projects/video-action-editing/paper/action_support_recovery.svg
new file mode 100644
index 0000000..83b4b68
--- /dev/null
+++ b/projects/video-action-editing/paper/action_support_recovery.svg
@@ -0,0 +1,305 @@
+
+
diff --git a/projects/video-action-editing/paper/data_generating_process.pdf b/projects/video-action-editing/paper/data_generating_process.pdf
new file mode 100644
index 0000000..9aa5dda
Binary files /dev/null and b/projects/video-action-editing/paper/data_generating_process.pdf differ
diff --git a/projects/video-action-editing/paper/data_generating_process.png b/projects/video-action-editing/paper/data_generating_process.png
new file mode 100644
index 0000000..5aa6408
Binary files /dev/null and b/projects/video-action-editing/paper/data_generating_process.png differ
diff --git a/projects/video-action-editing/paper/equations/action-condition.svg b/projects/video-action-editing/paper/equations/action-condition.svg
new file mode 100644
index 0000000..7af000f
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/action-condition.svg
@@ -0,0 +1,57 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/action-condition.tex b/projects/video-action-editing/paper/equations/action-condition.tex
new file mode 100644
index 0000000..9912839
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/action-condition.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 3_method.tex: \mathbf K(V_{\mathrm r})
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$\mathbf K(V_{\mathrm r})$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/action-margin.svg b/projects/video-action-editing/paper/equations/action-margin.svg
new file mode 100644
index 0000000..3d22bbf
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/action-margin.svg
@@ -0,0 +1,37 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/action-margin.tex b/projects/video-action-editing/paper/equations/action-margin.tex
new file mode 100644
index 0000000..e29b9a7
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/action-margin.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from appendix_b_unified_observations.tex: m_{\mathrm{act}}
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$m_{\mathrm{act}}$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/action-packet.svg b/projects/video-action-editing/paper/equations/action-packet.svg
new file mode 100644
index 0000000..8b838fe
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/action-packet.svg
@@ -0,0 +1,18 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/action-packet.tex b/projects/video-action-editing/paper/equations/action-packet.tex
new file mode 100644
index 0000000..28e2495
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/action-packet.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 3_method.tex: \mathbf K
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$\mathbf K$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/action-regime.svg b/projects/video-action-editing/paper/equations/action-regime.svg
new file mode 100644
index 0000000..f0c2dce
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/action-regime.svg
@@ -0,0 +1,18 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/action-regime.tex b/projects/video-action-editing/paper/equations/action-regime.tex
new file mode 100644
index 0000000..6bfb624
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/action-regime.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 2_theory.tex: u
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$u$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/action-trajectory.svg b/projects/video-action-editing/paper/equations/action-trajectory.svg
new file mode 100644
index 0000000..f37d627
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/action-trajectory.svg
@@ -0,0 +1,50 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/action-trajectory.tex b/projects/video-action-editing/paper/equations/action-trajectory.tex
new file mode 100644
index 0000000..05e0fbd
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/action-trajectory.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 2_theory.tex: \mathbf z^{\mathrm a}_{0:T}
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$\mathbf z^{\mathrm a}_{0:T}$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/adapter-update.svg b/projects/video-action-editing/paper/equations/adapter-update.svg
new file mode 100644
index 0000000..741c6e3
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/adapter-update.svg
@@ -0,0 +1,203 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/adapter-update.tex b/projects/video-action-editing/paper/equations/adapter-update.tex
new file mode 100644
index 0000000..d97cce4
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/adapter-update.tex
@@ -0,0 +1,12 @@
+% Extracted unchanged from 3_method.tex: eq:method-adapter
+\documentclass[border=4pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\begin{document}
+$\displaystyle
+\Delta \mathbf u=\alpha_k \mathbf W_{\mathrm{up}}\!\left[
+ (\mathbf W_u \mathbf u)\odot\tanh(\mathbf W_k \mathbf k)
+ +(\mathbf W_k \mathbf k)\odot\bigl(\mathbf 1+\tanh(\mathbf W_b \mathbf b)\bigr)\right].
+$
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/anchor-identification.svg b/projects/video-action-editing/paper/equations/anchor-identification.svg
new file mode 100644
index 0000000..3f986bd
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/anchor-identification.svg
@@ -0,0 +1,177 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/anchor-identification.tex b/projects/video-action-editing/paper/equations/anchor-identification.tex
new file mode 100644
index 0000000..fb69b5b
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/anchor-identification.tex
@@ -0,0 +1,12 @@
+% Extracted unchanged from 2_theory.tex: eq:anchor-identification
+\documentclass[border=4pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\begin{document}
+$\displaystyle
+E_{\mathrm a}(V)=h_{\mathrm a}(\mathbf z^{\mathrm a}),\qquad
+ P(h_{\mathrm a}(\mathbf z^{\mathrm a}))=\mu_k(\mathbf z^{\mathrm a})
+ \quad\text{almost surely},
+$
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/anchor-loss.svg b/projects/video-action-editing/paper/equations/anchor-loss.svg
new file mode 100644
index 0000000..f8cba98
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/anchor-loss.svg
@@ -0,0 +1,128 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/anchor-loss.tex b/projects/video-action-editing/paper/equations/anchor-loss.tex
new file mode 100644
index 0000000..ed8bfda
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/anchor-loss.tex
@@ -0,0 +1,13 @@
+% Extracted unchanged from 3_method.tex: eq:method-anchor-loss
+\documentclass[border=4pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\begin{document}
+$\displaystyle
+\mathcal L_{\mathrm{anchor}}
+ =\lambda_{\mathrm{obs}}\mathcal L_{\mathrm{obs}}
+ +\lambda_{\mathrm{view}}\mathcal L_{\mathrm{view}}
+ +\mathcal L_{\mathrm{reg}}.
+$
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/anchor-objective.svg b/projects/video-action-editing/paper/equations/anchor-objective.svg
new file mode 100644
index 0000000..cbd78d1
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/anchor-objective.svg
@@ -0,0 +1,303 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/anchor-objective.tex b/projects/video-action-editing/paper/equations/anchor-objective.tex
new file mode 100644
index 0000000..ac300fa
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/anchor-objective.tex
@@ -0,0 +1,15 @@
+% Extracted unchanged from 2_theory.tex: eq:anchor-objective
+\documentclass[border=4pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\begin{document}
+$\displaystyle
+\begin{aligned}
+ \mathcal J(E_{\mathrm a},P)
+ ={}&\lambda\,\underbrace{\E\|E_{\mathrm a}(V)-E_{\mathrm a}(V^+)\|^2}_{\mathcal L_{\mathrm{inv}}}\\
+ &+\underbrace{\E\|P(E_{\mathrm a}(V))-\varphi(V^+)\|_{\mathcal H_k}^2}_{\mathcal L_{\mathrm{pred}}},
+ \qquad\lambda>0 .
+ \end{aligned}
+$
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/anchor-views.svg b/projects/video-action-editing/paper/equations/anchor-views.svg
new file mode 100644
index 0000000..02e86cb
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/anchor-views.svg
@@ -0,0 +1,151 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/anchor-views.tex b/projects/video-action-editing/paper/equations/anchor-views.tex
new file mode 100644
index 0000000..6acf9eb
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/anchor-views.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 2_theory.tex: eq:controlled-anchor-views
+\documentclass[border=4pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\begin{document}
+$\displaystyle
+V=g_{\mathrm{ctrl}}(\mathbf z^{\mathrm c},\mathcal F_T(\mathbf w)),\qquad
+ V^+=g_{\mathrm{ctrl}}(\mathbf z^{\mathrm c+},\mathcal F_T(\mathbf w)).
+$
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/behavior-loss.svg b/projects/video-action-editing/paper/equations/behavior-loss.svg
new file mode 100644
index 0000000..0eb3213
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/behavior-loss.svg
@@ -0,0 +1,199 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/behavior-loss.tex b/projects/video-action-editing/paper/equations/behavior-loss.tex
new file mode 100644
index 0000000..542dce8
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/behavior-loss.tex
@@ -0,0 +1,13 @@
+% Extracted unchanged from 3_method.tex: eq:method-behavior-loss
+\documentclass[border=4pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\begin{document}
+$\displaystyle
+\ell_{\mathrm{act},k}=1-\gamma_{+,k}
+ -\lambda_{\mathrm{disc}}\log
+ \frac{\exp(\gamma_{+,k}/\tau)}
+ {\sum_{j\in\mathcal K}\exp(\gamma_{j,k}/\tau)}
+$
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/behavior-similarity.svg b/projects/video-action-editing/paper/equations/behavior-similarity.svg
new file mode 100644
index 0000000..f3cf28d
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/behavior-similarity.svg
@@ -0,0 +1,126 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/behavior-similarity.tex b/projects/video-action-editing/paper/equations/behavior-similarity.tex
new file mode 100644
index 0000000..b116993
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/behavior-similarity.tex
@@ -0,0 +1,10 @@
+% Extracted unchanged from 3_method.tex: \gamma_{j,k}=\operatorname{sim}(e_{\mathrm J}(\widehat V_{0\mid k}),e_{\mathrm J}(V_j))
+\documentclass[border=4pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\begin{document}
+$\displaystyle
+\gamma_{j,k}=\operatorname{sim}(e_{\mathrm J}(\widehat V_{0\mid k}),e_{\mathrm J}(V_j))
+$
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/content-code.svg b/projects/video-action-editing/paper/equations/content-code.svg
new file mode 100644
index 0000000..6e23f14
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/content-code.svg
@@ -0,0 +1,29 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/content-code.tex b/projects/video-action-editing/paper/equations/content-code.tex
new file mode 100644
index 0000000..e2261b6
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/content-code.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 2_theory.tex: \mathbf z^{\mathrm c}
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$\mathbf z^{\mathrm c}$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/demonstration-video.svg b/projects/video-action-editing/paper/equations/demonstration-video.svg
new file mode 100644
index 0000000..cd13d6e
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/demonstration-video.svg
@@ -0,0 +1,29 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/demonstration-video.tex b/projects/video-action-editing/paper/equations/demonstration-video.tex
new file mode 100644
index 0000000..8af845a
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/demonstration-video.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 3_method.tex: V_{\mathrm r}
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$V_{\mathrm r}$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/generated-token.svg b/projects/video-action-editing/paper/equations/generated-token.svg
new file mode 100644
index 0000000..6ea97b1
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/generated-token.svg
@@ -0,0 +1,18 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/generated-token.tex b/projects/video-action-editing/paper/equations/generated-token.tex
new file mode 100644
index 0000000..eb76e51
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/generated-token.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 3_method.tex: \mathbf u
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$\mathbf u$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/identification-map.svg b/projects/video-action-editing/paper/equations/identification-map.svg
new file mode 100644
index 0000000..5234c07
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/identification-map.svg
@@ -0,0 +1,29 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/identification-map.tex b/projects/video-action-editing/paper/equations/identification-map.tex
new file mode 100644
index 0000000..412f9c2
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/identification-map.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 2_theory.tex: h_{\mathrm a}
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$h_{\mathrm a}$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/kernel-mean.svg b/projects/video-action-editing/paper/equations/kernel-mean.svg
new file mode 100644
index 0000000..ab8354b
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/kernel-mean.svg
@@ -0,0 +1,29 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/kernel-mean.tex b/projects/video-action-editing/paper/equations/kernel-mean.tex
new file mode 100644
index 0000000..c83c81e
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/kernel-mean.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 2_theory.tex: \mu_k
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$\mu_k$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/manifest.json b/projects/video-action-editing/paper/equations/manifest.json
new file mode 100644
index 0000000..bf142af
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/manifest.json
@@ -0,0 +1,400 @@
+[
+ {
+ "name": "anchor-views",
+ "source": "/paper/source/2_theory.tex",
+ "source_sha256": "cea2d74371d5ebc780faf1715c3ed386bbc5263309f84754389b2057df4f35d7",
+ "extraction": "equation",
+ "selector": "eq:controlled-anchor-views",
+ "body": "V=g_{\\mathrm{ctrl}}(\\mathbf z^{\\mathrm c},\\mathcal F_T(\\mathbf w)),\\qquad\n V^+=g_{\\mathrm{ctrl}}(\\mathbf z^{\\mathrm c+},\\mathcal F_T(\\mathbf w)).",
+ "inline": false,
+ "single_line": true,
+ "latex": "/paper/equations/anchor-views.tex",
+ "svg": "/paper/equations/anchor-views.svg",
+ "svg_sha256": "dc739ae5d698fd321f344453458481ed2ce371307a83d50b4a4e7d13e108496e",
+ "label": "eq:controlled-anchor-views"
+ },
+ {
+ "name": "anchor-objective",
+ "source": "/paper/source/2_theory.tex",
+ "source_sha256": "cea2d74371d5ebc780faf1715c3ed386bbc5263309f84754389b2057df4f35d7",
+ "extraction": "equation",
+ "selector": "eq:anchor-objective",
+ "body": "\\begin{aligned}\n \\mathcal J(E_{\\mathrm a},P)\n ={}&\\lambda\\,\\underbrace{\\E\\|E_{\\mathrm a}(V)-E_{\\mathrm a}(V^+)\\|^2}_{\\mathcal L_{\\mathrm{inv}}}\\\\\n &+\\underbrace{\\E\\|P(E_{\\mathrm a}(V))-\\varphi(V^+)\\|_{\\mathcal H_k}^2}_{\\mathcal L_{\\mathrm{pred}}},\n \\qquad\\lambda>0 .\n \\end{aligned}",
+ "inline": false,
+ "single_line": false,
+ "latex": "/paper/equations/anchor-objective.tex",
+ "svg": "/paper/equations/anchor-objective.svg",
+ "svg_sha256": "11d777ce785478829a7a7a4b73428b6f3c217d5dda22cd22e7645930283796fa",
+ "label": "eq:anchor-objective"
+ },
+ {
+ "name": "anchor-identification",
+ "source": "/paper/source/2_theory.tex",
+ "source_sha256": "cea2d74371d5ebc780faf1715c3ed386bbc5263309f84754389b2057df4f35d7",
+ "extraction": "equation",
+ "selector": "eq:anchor-identification",
+ "body": "E_{\\mathrm a}(V)=h_{\\mathrm a}(\\mathbf z^{\\mathrm a}),\\qquad\n P(h_{\\mathrm a}(\\mathbf z^{\\mathrm a}))=\\mu_k(\\mathbf z^{\\mathrm a})\n \\quad\\text{almost surely},",
+ "inline": false,
+ "single_line": true,
+ "latex": "/paper/equations/anchor-identification.tex",
+ "svg": "/paper/equations/anchor-identification.svg",
+ "svg_sha256": "b2ed758116dc9d6b8b95933f859daf7893acc7ed5c59972cd96c740adf7f8e5d",
+ "label": "eq:anchor-identification"
+ },
+ {
+ "name": "temporal-observations",
+ "source": "/paper/source/2_theory.tex",
+ "source_sha256": "cea2d74371d5ebc780faf1715c3ed386bbc5263309f84754389b2057df4f35d7",
+ "extraction": "display",
+ "selector": "p_t=p_0+t v_0+\\tfrac12t^2u",
+ "body": "p_t=p_0+t v_0+\\tfrac12t^2u,\\qquad \\mathbf x_t=\\rho(c,p_t),",
+ "inline": false,
+ "single_line": true,
+ "latex": "/paper/equations/temporal-observations.tex",
+ "svg": "/paper/equations/temporal-observations.svg",
+ "svg_sha256": "8a377fe29884d789124c2d6deb4df7d14e17a0d1ed40a7aecb8297a0c8603e6a"
+ },
+ {
+ "name": "anchor-loss",
+ "source": "/paper/source/3_method.tex",
+ "source_sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "extraction": "equation",
+ "selector": "eq:method-anchor-loss",
+ "body": "\\mathcal L_{\\mathrm{anchor}}\n =\\lambda_{\\mathrm{obs}}\\mathcal L_{\\mathrm{obs}}\n +\\lambda_{\\mathrm{view}}\\mathcal L_{\\mathrm{view}}\n +\\mathcal L_{\\mathrm{reg}}.",
+ "inline": false,
+ "single_line": true,
+ "latex": "/paper/equations/anchor-loss.tex",
+ "svg": "/paper/equations/anchor-loss.svg",
+ "svg_sha256": "0ea8b09e8b5afb187bb769a5f427dbcef00ae3d15a70bd4a6d26cdb360c22b82",
+ "label": "eq:method-anchor-loss"
+ },
+ {
+ "name": "adapter-update",
+ "source": "/paper/source/3_method.tex",
+ "source_sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "extraction": "equation",
+ "selector": "eq:method-adapter",
+ "body": "\\Delta \\mathbf u=\\alpha_k \\mathbf W_{\\mathrm{up}}\\!\\left[\n (\\mathbf W_u \\mathbf u)\\odot\\tanh(\\mathbf W_k \\mathbf k)\n +(\\mathbf W_k \\mathbf k)\\odot\\bigl(\\mathbf 1+\\tanh(\\mathbf W_b \\mathbf b)\\bigr)\\right].",
+ "inline": false,
+ "single_line": true,
+ "latex": "/paper/equations/adapter-update.tex",
+ "svg": "/paper/equations/adapter-update.svg",
+ "svg_sha256": "8645a29c0f488e8219dd53deecaba8a0e90d4a29aa6ca28bb3a87411e0260550",
+ "label": "eq:method-adapter"
+ },
+ {
+ "name": "behavior-similarity",
+ "source": "/paper/source/3_method.tex",
+ "source_sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "extraction": "inline",
+ "selector": "\\gamma_{j,k}=\\operatorname{sim}(e_{\\mathrm J}(\\widehat V_{0\\mid k}),e_{\\mathrm J}(V_j))",
+ "body": "\\gamma_{j,k}=\\operatorname{sim}(e_{\\mathrm J}(\\widehat V_{0\\mid k}),e_{\\mathrm J}(V_j))",
+ "inline": false,
+ "single_line": true,
+ "latex": "/paper/equations/behavior-similarity.tex",
+ "svg": "/paper/equations/behavior-similarity.svg",
+ "svg_sha256": "e431339494a4833e08b41719a8d0df9e8d6bf0affc02f626488cf430c3c64123"
+ },
+ {
+ "name": "behavior-loss",
+ "source": "/paper/source/3_method.tex",
+ "source_sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "extraction": "equation",
+ "selector": "eq:method-behavior-loss",
+ "body": "\\ell_{\\mathrm{act},k}=1-\\gamma_{+,k}\n -\\lambda_{\\mathrm{disc}}\\log\n \\frac{\\exp(\\gamma_{+,k}/\\tau)}\n {\\sum_{j\\in\\mathcal K}\\exp(\\gamma_{j,k}/\\tau)}",
+ "inline": false,
+ "single_line": true,
+ "latex": "/paper/equations/behavior-loss.tex",
+ "svg": "/paper/equations/behavior-loss.svg",
+ "svg_sha256": "3fc4bb8343f4bfaeb9d2c000f827ad8843235ef1327f6729a7529e0da914a04b",
+ "label": "eq:method-behavior-loss"
+ },
+ {
+ "name": "multiplicative-map",
+ "source": "/paper/source/appendix_e_synthetic_protocol.tex",
+ "source_sha256": "4a28dec0bbbd86de384ff2471c30d3f3e2e188855eb650553520f7a698127dad",
+ "extraction": "inline",
+ "selector": "\\rho(c,p)=(cp,c)",
+ "body": "\\rho(c,p)=(cp,c)",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/multiplicative-map.tex",
+ "svg": "/paper/equations/multiplicative-map.svg",
+ "svg_sha256": "82bf133a32ea4ef69d5d5bc90ebb593fe2bb86309e7545ff73bfa6a0d31a3ce2",
+ "width_em": 6.5993,
+ "height_em": 0.75,
+ "depth_em": 0.25
+ },
+ {
+ "name": "zero-conditional-mean",
+ "source": "/paper/source/appendix_e_synthetic_protocol.tex",
+ "source_sha256": "4a28dec0bbbd86de384ff2471c30d3f3e2e188855eb650553520f7a698127dad",
+ "extraction": "inline",
+ "selector": "\\E[V\\mid\\mathbf z^{\\mathrm a}]=\\mathbf 0",
+ "body": "\\E[V\\mid\\mathbf z^{\\mathrm a}]=\\mathbf 0",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/zero-conditional-mean.tex",
+ "svg": "/paper/equations/zero-conditional-mean.svg",
+ "svg_sha256": "49d98decb307591099d1c5e646fa0e6356ee5f05a2d98fd3ba127d1ab5e5d795",
+ "width_em": 5.72914,
+ "height_em": 0.75,
+ "depth_em": 0.25
+ },
+ {
+ "name": "content-code",
+ "source": "/paper/source/2_theory.tex",
+ "source_sha256": "cea2d74371d5ebc780faf1715c3ed386bbc5263309f84754389b2057df4f35d7",
+ "extraction": "inline",
+ "selector": "\\mathbf z^{\\mathrm c}",
+ "body": "\\mathbf z^{\\mathrm c}",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/content-code.tex",
+ "svg": "/paper/equations/content-code.svg",
+ "svg_sha256": "860f090a6343d40afb8aad52269a00f967e374cca2173ee6fd319bc554f31b50",
+ "width_em": 0.91667,
+ "height_em": 0.66428,
+ "depth_em": 0.0
+ },
+ {
+ "name": "action-trajectory",
+ "source": "/paper/source/2_theory.tex",
+ "source_sha256": "cea2d74371d5ebc780faf1715c3ed386bbc5263309f84754389b2057df4f35d7",
+ "extraction": "fragment",
+ "selector": "\\mathbf z^{\\mathrm a}_{0:T}",
+ "body": "\\mathbf z^{\\mathrm a}_{0:T}",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/action-trajectory.tex",
+ "svg": "/paper/equations/action-trajectory.svg",
+ "svg_sha256": "9757c94d72829e4c42d310b809c1ee9efb6a07d1535c581f14bb5399014ea8e9",
+ "width_em": 1.76608,
+ "height_em": 0.66428,
+ "depth_em": 0.27543
+ },
+ {
+ "name": "action-regime",
+ "source": "/paper/source/2_theory.tex",
+ "source_sha256": "cea2d74371d5ebc780faf1715c3ed386bbc5263309f84754389b2057df4f35d7",
+ "extraction": "inline",
+ "selector": "u",
+ "body": "u",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/action-regime.tex",
+ "svg": "/paper/equations/action-regime.svg",
+ "svg_sha256": "0c9f2dc8308bdcd5be55a457f7f94837efc09d16511ddb2103954d8b8fbbf15a",
+ "width_em": 0.57246,
+ "height_em": 0.43055,
+ "depth_em": 0.0
+ },
+ {
+ "name": "identification-map",
+ "source": "/paper/source/2_theory.tex",
+ "source_sha256": "cea2d74371d5ebc780faf1715c3ed386bbc5263309f84754389b2057df4f35d7",
+ "extraction": "inline",
+ "selector": "h_{\\mathrm a}",
+ "body": "h_{\\mathrm a}",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/identification-map.tex",
+ "svg": "/paper/equations/identification-map.svg",
+ "svg_sha256": "9d7de8f8db084c9bd4efcc4f32b85d5f930d6f01237fd12b1747b130f70bb51b",
+ "width_em": 1.02477,
+ "height_em": 0.69444,
+ "depth_em": 0.15
+ },
+ {
+ "name": "kernel-mean",
+ "source": "/paper/source/2_theory.tex",
+ "source_sha256": "cea2d74371d5ebc780faf1715c3ed386bbc5263309f84754389b2057df4f35d7",
+ "extraction": "fragment",
+ "selector": "\\mu_k",
+ "body": "\\mu_k",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/kernel-mean.tex",
+ "svg": "/paper/equations/kernel-mean.svg",
+ "svg_sha256": "9f14a2e57615009e6fdd597069463d058832c58403f546f24ef25980b8a640a6",
+ "width_em": 1.09457,
+ "height_em": 0.43055,
+ "depth_em": 0.19444
+ },
+ {
+ "name": "demonstration-video",
+ "source": "/paper/source/3_method.tex",
+ "source_sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "extraction": "inline",
+ "selector": "V_{\\mathrm r}",
+ "body": "V_{\\mathrm r}",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/demonstration-video.tex",
+ "svg": "/paper/equations/demonstration-video.svg",
+ "svg_sha256": "2c79e2497e6a249486e9308da253a796d8650d5b12e00955eb416284c1edfcd0",
+ "width_em": 0.94583,
+ "height_em": 0.68333,
+ "depth_em": 0.15
+ },
+ {
+ "name": "action-condition",
+ "source": "/paper/source/3_method.tex",
+ "source_sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "extraction": "fragment",
+ "selector": "\\mathbf K(V_{\\mathrm r})",
+ "body": "\\mathbf K(V_{\\mathrm r})",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/action-condition.tex",
+ "svg": "/paper/equations/action-condition.svg",
+ "svg_sha256": "37062d3c825580b40956b7a46912e7e4a00c1e8434ee0102be0464429aa586f3",
+ "width_em": 2.625,
+ "height_em": 0.75,
+ "depth_em": 0.25
+ },
+ {
+ "name": "action-packet",
+ "source": "/paper/source/3_method.tex",
+ "source_sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "extraction": "inline",
+ "selector": "\\mathbf K",
+ "body": "\\mathbf K",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/action-packet.tex",
+ "svg": "/paper/equations/action-packet.svg",
+ "svg_sha256": "8d4a716fdb5d52503f43ceef63577d5500d944b5f23a1c725c0dcf1f891127de",
+ "width_em": 0.90138,
+ "height_em": 0.68611,
+ "depth_em": 0.0
+ },
+ {
+ "name": "generated-token",
+ "source": "/paper/source/3_method.tex",
+ "source_sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "extraction": "inline",
+ "selector": "\\mathbf u",
+ "body": "\\mathbf u",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/generated-token.tex",
+ "svg": "/paper/equations/generated-token.svg",
+ "svg_sha256": "8fe941b5fa7f85a06309001efa7e74452fbf883a95c618d84ed2e20a324d2989",
+ "width_em": 0.63888,
+ "height_em": 0.44444,
+ "depth_em": 0.0
+ },
+ {
+ "name": "source-context",
+ "source": "/paper/source/3_method.tex",
+ "source_sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "extraction": "inline",
+ "selector": "\\mathbf b",
+ "body": "\\mathbf b",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/source-context.tex",
+ "svg": "/paper/equations/source-context.svg",
+ "svg_sha256": "bbe0b3972c9a3ba28bff05a7983378cf5c8c4bb670804f43788dd8f641381b08",
+ "width_em": 0.63888,
+ "height_em": 0.69444,
+ "depth_em": 0.0
+ },
+ {
+ "name": "sampling-gate",
+ "source": "/paper/source/3_method.tex",
+ "source_sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "extraction": "inline",
+ "selector": "\\alpha_k",
+ "body": "\\alpha_k",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/sampling-gate.tex",
+ "svg": "/paper/equations/sampling-gate.svg",
+ "svg_sha256": "1f5f2559e0a3445cd88a3525eaa8389056e2372090a8ea1d980502b061d2afd4",
+ "width_em": 1.13172,
+ "height_em": 0.43055,
+ "depth_em": 0.15
+ },
+ {
+ "name": "action-margin",
+ "source": "/paper/source/appendix_b_unified_observations.tex",
+ "source_sha256": "95a7866b782ff54a2a6c0c1bfbdfe0f1c59eb0dfcb2fc6af4e72ace1a857337b",
+ "extraction": "inline",
+ "selector": "m_{\\mathrm{act}}",
+ "body": "m_{\\mathrm{act}}",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/action-margin.tex",
+ "svg": "/paper/equations/action-margin.svg",
+ "svg_sha256": "e21aac7853ff1dd9736f96d0ecbcc711d3c67b5a312b9a78d30fc15bf6bc6c27",
+ "width_em": 1.99469,
+ "height_em": 0.43055,
+ "depth_em": 0.15
+ },
+ {
+ "name": "r-squared",
+ "source": "/paper/source/appendix_e_synthetic_protocol.tex",
+ "source_sha256": "4a28dec0bbbd86de384ff2471c30d3f3e2e188855eb650553520f7a698127dad",
+ "extraction": "inline",
+ "selector": "R^2",
+ "body": "R^2",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/r-squared.tex",
+ "svg": "/paper/equations/r-squared.svg",
+ "svg_sha256": "63cbbee88673c779f5921f5b31bd90d2431ec8ac8519a8ed046a1fd823159575",
+ "width_em": 1.21563,
+ "height_em": 0.814,
+ "depth_em": 0.0
+ },
+ {
+ "name": "midpoint-position",
+ "source": "/paper/source/appendix_e_synthetic_protocol.tex",
+ "source_sha256": "4a28dec0bbbd86de384ff2471c30d3f3e2e188855eb650553520f7a698127dad",
+ "extraction": "inline",
+ "selector": "p",
+ "body": "p",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/midpoint-position.tex",
+ "svg": "/paper/equations/midpoint-position.svg",
+ "svg_sha256": "47d675d737b942475834c217d7d3cd9a794c9bb85ce8773fe339789ef271c568",
+ "width_em": 0.50313,
+ "height_em": 0.43055,
+ "depth_em": 0.19444
+ },
+ {
+ "name": "velocity",
+ "source": "/paper/source/appendix_e_synthetic_protocol.tex",
+ "source_sha256": "4a28dec0bbbd86de384ff2471c30d3f3e2e188855eb650553520f7a698127dad",
+ "extraction": "inline",
+ "selector": "v",
+ "body": "v",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/velocity.tex",
+ "svg": "/paper/equations/velocity.svg",
+ "svg_sha256": "ad5d5b38263b131822c61bc43116c70dfa0dcbd6ab8f727795ec14917e9bd67b",
+ "width_em": 0.5206,
+ "height_em": 0.43055,
+ "depth_em": 0.0
+ },
+ {
+ "name": "three-frames",
+ "source": "/paper/source/appendix_e_synthetic_protocol.tex",
+ "source_sha256": "4a28dec0bbbd86de384ff2471c30d3f3e2e188855eb650553520f7a698127dad",
+ "extraction": "inline",
+ "selector": "T=3",
+ "body": "T=3",
+ "inline": true,
+ "single_line": true,
+ "latex": "/paper/equations/three-frames.tex",
+ "svg": "/paper/equations/three-frames.svg",
+ "svg_sha256": "dea0511da14cb6ae10810722650d641ac869aac227eb6a98f3a5e77781f812fc",
+ "width_em": 2.55659,
+ "height_em": 0.68333,
+ "depth_em": 0.0
+ }
+]
diff --git a/projects/video-action-editing/paper/equations/midpoint-position.svg b/projects/video-action-editing/paper/equations/midpoint-position.svg
new file mode 100644
index 0000000..f123ded
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/midpoint-position.svg
@@ -0,0 +1,18 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/midpoint-position.tex b/projects/video-action-editing/paper/equations/midpoint-position.tex
new file mode 100644
index 0000000..4b01415
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/midpoint-position.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from appendix_e_synthetic_protocol.tex: p
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$p$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/multiplicative-map.svg b/projects/video-action-editing/paper/equations/multiplicative-map.svg
new file mode 100644
index 0000000..de076e7
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/multiplicative-map.svg
@@ -0,0 +1,101 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/multiplicative-map.tex b/projects/video-action-editing/paper/equations/multiplicative-map.tex
new file mode 100644
index 0000000..8c5446d
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/multiplicative-map.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from appendix_e_synthetic_protocol.tex: \rho(c,p)=(cp,c)
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$\rho(c,p)=(cp,c)$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/r-squared.svg b/projects/video-action-editing/paper/equations/r-squared.svg
new file mode 100644
index 0000000..4d88260
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/r-squared.svg
@@ -0,0 +1,24 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/r-squared.tex b/projects/video-action-editing/paper/equations/r-squared.tex
new file mode 100644
index 0000000..e949f71
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/r-squared.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from appendix_e_synthetic_protocol.tex: R^2
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$R^2$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/sampling-gate.svg b/projects/video-action-editing/paper/equations/sampling-gate.svg
new file mode 100644
index 0000000..6c4d952
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/sampling-gate.svg
@@ -0,0 +1,29 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/sampling-gate.tex b/projects/video-action-editing/paper/equations/sampling-gate.tex
new file mode 100644
index 0000000..00824b6
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/sampling-gate.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 3_method.tex: \alpha_k
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$\alpha_k$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/source-context.svg b/projects/video-action-editing/paper/equations/source-context.svg
new file mode 100644
index 0000000..91230c7
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/source-context.svg
@@ -0,0 +1,18 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/source-context.tex b/projects/video-action-editing/paper/equations/source-context.tex
new file mode 100644
index 0000000..e4dc823
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/source-context.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from 3_method.tex: \mathbf b
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$\mathbf b$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/temporal-observations.svg b/projects/video-action-editing/paper/equations/temporal-observations.svg
new file mode 100644
index 0000000..d09489c
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/temporal-observations.svg
@@ -0,0 +1,131 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/temporal-observations.tex b/projects/video-action-editing/paper/equations/temporal-observations.tex
new file mode 100644
index 0000000..cfb800d
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/temporal-observations.tex
@@ -0,0 +1,10 @@
+% Extracted unchanged from 2_theory.tex: p_t=p_0+t v_0+\tfrac12t^2u
+\documentclass[border=4pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\begin{document}
+$\displaystyle
+p_t=p_0+t v_0+\tfrac12t^2u,\qquad \mathbf x_t=\rho(c,p_t),
+$
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/three-frames.svg b/projects/video-action-editing/paper/equations/three-frames.svg
new file mode 100644
index 0000000..31b81c1
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/three-frames.svg
@@ -0,0 +1,35 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/three-frames.tex b/projects/video-action-editing/paper/equations/three-frames.tex
new file mode 100644
index 0000000..c1b516e
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/three-frames.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from appendix_e_synthetic_protocol.tex: T=3
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$T=3$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/velocity.svg b/projects/video-action-editing/paper/equations/velocity.svg
new file mode 100644
index 0000000..e295f90
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/velocity.svg
@@ -0,0 +1,18 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/velocity.tex b/projects/video-action-editing/paper/equations/velocity.tex
new file mode 100644
index 0000000..436f9c0
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/velocity.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from appendix_e_synthetic_protocol.tex: v
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$v$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/equations/zero-conditional-mean.svg b/projects/video-action-editing/paper/equations/zero-conditional-mean.svg
new file mode 100644
index 0000000..c46ab5a
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/zero-conditional-mean.svg
@@ -0,0 +1,96 @@
+
+
diff --git a/projects/video-action-editing/paper/equations/zero-conditional-mean.tex b/projects/video-action-editing/paper/equations/zero-conditional-mean.tex
new file mode 100644
index 0000000..9c049ca
--- /dev/null
+++ b/projects/video-action-editing/paper/equations/zero-conditional-mean.tex
@@ -0,0 +1,11 @@
+% Extracted unchanged from appendix_e_synthetic_protocol.tex: \E[V\mid\mathbf z^{\mathrm a}]=\mathbf 0
+\documentclass[border=0pt]{standalone}
+\usepackage{times}
+\usepackage{amsmath,amsfonts,bm}
+\newcommand{\E}{\mathbb{E}}
+\newsavebox{\equationbox}
+\begin{document}%
+\savebox{\equationbox}{$\E[V\mid\mathbf z^{\mathrm a}]=\mathbf 0$}%
+\typeout{MATHBOX:\the\wd\equationbox,\the\ht\equationbox,\the\dp\equationbox}%
+\usebox{\equationbox}%
+\end{document}
diff --git a/projects/video-action-editing/paper/source/2_theory.tex b/projects/video-action-editing/paper/source/2_theory.tex
new file mode 100644
index 0000000..21f34ab
--- /dev/null
+++ b/projects/video-action-editing/paper/source/2_theory.tex
@@ -0,0 +1,191 @@
+\section{Theory}
+\label{sec:theory}
+We model action editing with a latent video-generating process, establish
+when paired-view learning identifies action anchors, and explain their role
+in demonstration-guided editing.
+
+\subsection{Problem formulation}
+\label{subsec:dgp-overview}
+
+For a video $V=(\mathbf x_1,\ldots,\mathbf x_T)$, let $\mathbf z^{\mathrm c}$
+denote persistent content and $\mathbf z^{\mathrm a}_t$ its action state.
+For a controlled description with physical context fixed, the generating process is
+\begin{equation}
+ \begin{aligned}
+ \mathbf z^{\mathrm c}&\sim p_{\mathrm c},\qquad (u,\mathbf z^{\mathrm a}_0,\boldsymbol\epsilon_{1:T})\sim p_{\mathrm{dyn}},\\
+ \mathbf z^{\mathrm a}_t&=f(\mathbf z^{\mathrm a}_{t-1},u,\boldsymbol\epsilon_t),\qquad
+ V=g_{\mathrm{ctrl}}(\mathbf z^{\mathrm c},\mathbf z^{\mathrm a}_{0:T}).
+ \end{aligned}
+ \label{eq:dgp-controlled}
+\end{equation}
+Here, $u$ specifies the action regime (e.g., raising or lowering an arm),
+and $\boldsymbol\epsilon_t$ captures stochastic variation in the state
+transition $f$. The observation map $g_{\mathrm{ctrl}}$ maps content and
+the resulting action trajectory to a video. $p_{\mathrm c}$ is the content
+distribution, while $p_{\mathrm{dyn}}$ is the joint distribution of $u$,
+the initial state $\mathbf z^{\mathrm a}_0$, and $\boldsymbol\epsilon_{1:T}$.
+The action block is $\mathbf z^{\mathrm a}:=\mathbf z^{\mathrm a}_{0:T}$.
+Content is independent of the action-generating variables
+(Appendix~\ref{sec:controlled-transfer}). For source
+$V_{\mathrm s}=g_{\mathrm{ctrl}}(\mathbf z^{\mathrm c}_{\mathrm s},\mathbf z^{\mathrm a}_{\mathrm s})$ and demonstration
+$V_{\mathrm r}=g_{\mathrm{ctrl}}(\mathbf z^{\mathrm c}_{\mathrm r},\mathbf z^{\mathrm a\star})$, the desired edit is
+\begin{equation}
+ V^\star=g_{\mathrm{ctrl}}(\mathbf z^{\mathrm c}_{\mathrm s},\mathbf z^{\mathrm a\star}).
+ \label{eq:dgp-counterfactual}
+\end{equation}
+Action-preserving pairs vary appearance at a fixed trajectory
+(Figure~\ref{fig:controlled-dgp}). We use this relation to study when
+an action representation can be recovered independently of appearance.
+
+\begin{figure}[!htbp]
+ \centering
+ \includegraphics[width=\linewidth]{data_generating_process.pdf}
+ \caption{Controlled video generation and content--action recombination within a fixed physical context.
+ Content $\mathbf z^{\mathrm c}$ and action $\mathbf z^{\mathrm a}_{0:T}$ jointly determine the video
+ through the renderer $g_{\mathrm{ctrl}}$ (left).
+ Rows share content and columns share action; the desired edit combines source content
+ $\mathbf z^{\mathrm c}_{\mathrm s}$ with demonstrated action $\mathbf z^{\mathrm a\star}$ (right).}
+ \label{fig:controlled-dgp}
+\end{figure}
+\subsection{Learning identifiable action anchors}
+\label{subsec:anchor-learning}
+\label{subsec:anchor-identification}
+
+A demonstration supplies temporal behavior together with its visual content.
+The learning problem is to retain the former across changes in the latter.
+Let $\mathbf w$ collect the initial state, regime, and innovations in
+Eq.~\ref{eq:dgp-controlled}, and let $\mathcal F_T(\mathbf w)=\mathbf z^{\mathrm a}_{0:T}$
+be its unrolled trajectory. An action-preserving pair holds $\mathbf w$ fixed:
+\begin{equation}
+ V=g_{\mathrm{ctrl}}(\mathbf z^{\mathrm c},\mathcal F_T(\mathbf w)),\qquad
+ V^+=g_{\mathrm{ctrl}}(\mathbf z^{\mathrm c+},\mathcal F_T(\mathbf w)).
+ \label{eq:controlled-anchor-views}
+\end{equation}
+The entire realized trajectory is shared, including timing and interactions;
+an action label alone does not specify this relation. These are
+representation-learning pairs; the source and target-action demonstration
+may depict different actions.
+
+\begin{definition}[Blockwise identification]
+\label{def:anchor-identification}
+An encoder $E_{\mathrm a}:\mathcal V\to\R^{d_{\mathrm a}}$ identifies the action block if
+$E_{\mathrm a}(V)=h_{\mathrm a}(\mathbf z^{\mathrm a})$ almost surely,
+where $h_{\mathrm a}$ is injective on a full-measure subset of the action support.
+Coordinates may mix within the block.
+\end{definition}
+
+\Needspace{13\baselineskip}
+\begin{condition}[Nonlinear action identification]
+\label{cond:anchor-identification}
+\leavevmode
+
+\par\noindent\textup{(i)} \textbf{Regularity.}
+The supports $\mathcal W$ and $\mathcal Z^{\mathrm c}$ are compact, and
+$f$ and $g_{\mathrm{ctrl}}$ are continuous. Write
+$\mathcal A_T=\mathcal F_T(\mathcal W)$ for the supported trajectory family.
+
+\par\noindent\textup{(ii)} \textbf{Appearance variation.}
+The two content variables in Eq.~\ref{eq:controlled-anchor-views} are
+independent draws from $p_{\mathrm c}$, independently of $\mathbf w$,
+and $p_{\mathrm c}$ has full support on $\mathcal Z^{\mathrm c}$.
+
+\par\noindent\textup{(iii)} \textbf{Temporal observability.}
+Two distinct trajectories in $\mathcal A_T$ cannot generate the same
+complete video under any supported content choices.
+\end{condition}
+Part (iii) concerns the ordered video, not each frame separately.
+An injective video-generating map suffices, but recovering content is
+unnecessary. The observation map may mix action and appearance nonlinearly;
+no additive feature decomposition is imposed.
+
+\paragraph{Population learning problem.}
+Agreement alone permits collapse, while predicting an ordinary feature
+mean can erase differences between actions. We therefore analyze prediction
+of the action-conditional video distribution. Let $k$ be a bounded,
+continuous characteristic kernel on ordered videos, with feature map
+$\varphi(V)=k(V,\cdot)$ in its reproducing kernel Hilbert space $\mathcal H_k$.
+``Characteristic'' means that the mean embedding uniquely determines a
+probability distribution; a Gaussian kernel on vectorized videos is one
+choice~\citep{sriperumbudur2010hilbert}. Consider
+\begin{equation}
+ \begin{aligned}
+ \mathcal J(E_{\mathrm a},P)
+ ={}&\lambda\,\underbrace{\E\|E_{\mathrm a}(V)-E_{\mathrm a}(V^+)\|^2}_{\mathcal L_{\mathrm{inv}}}\\
+ &+\underbrace{\E\|P(E_{\mathrm a}(V))-\varphi(V^+)\|_{\mathcal H_k}^2}_{\mathcal L_{\mathrm{pred}}},
+ \qquad\lambda>0 .
+ \end{aligned}
+ \label{eq:anchor-objective}
+\end{equation}
+We optimize over all continuous encoders and measurable predictors
+with finite risk. The code dimension must admit continuous
+injective coordinates of $\mathcal A_T$; the full trajectory always
+provides such coordinates. This criterion uses observed view pairs
+without action-coordinate labels.
+
+\begin{theorem}[Blockwise identification by paired-view learning]
+\label{thm:anchor-identification}
+Under Eq.~\ref{eq:controlled-anchor-views} and
+Condition~\ref{cond:anchor-identification}, the population class above
+attains the minimum of Eq.~\ref{eq:anchor-objective}.
+Every global minimizer satisfies
+\begin{equation}
+ E_{\mathrm a}(V)=h_{\mathrm a}(\mathbf z^{\mathrm a}),\qquad
+ P(h_{\mathrm a}(\mathbf z^{\mathrm a}))=\mu_k(\mathbf z^{\mathrm a})
+ \quad\text{almost surely},
+ \label{eq:anchor-identification}
+\end{equation}
+where $\mu_k(\mathbf a)=\E[\varphi(V)\mid\mathbf z^{\mathrm a}=\mathbf a]$
+and $h_{\mathrm a}$ is injective on a full-measure action subset.
+\end{theorem}
+
+\paragraph{Proof sketch.}
+Temporal observability and compactness give a continuous action readout
+from the whole video. Distinct actions induce distributions supported on
+disjoint sets of videos, so their characteristic embeddings differ.
+These facts construct an invariant encoder and predictor attaining the
+irreducible prediction variance. Conditional independence decomposes the
+objective into this variance and two nonnegative errors. Every global
+optimum therefore eliminates appearance variation while retaining distinct
+action distributions (Appendix~\ref{subsec:nonlinear-anchor-proof}).
+
+\subsection{Temporal information and action transfer}
+\label{subsec:anchor-realization}
+
+Consider position--velocity dynamics with constant acceleration $u$ and
+nonlinear frame observations
+\[
+ p_t=p_0+t v_0+\tfrac12t^2u,\qquad \mathbf x_t=\rho(c,p_t),
+\]
+where $\rho$ is any continuous invertible map with continuous inverse
+on the supported inputs. Three distinct observation times determine
+$(p_0,v_0,u)$ and the full trajectory, whereas two times leave an ambiguity.
+Thus single-frame recovery of the action block is unnecessary, even when
+appearance and position interact nonlinearly
+(Appendix~\ref{subsec:nonlinear-temporal-example}).
+
+For additive features, the prediction target can be simplified. On the
+supported content--trajectory pairs, let
+\[
+ \Phi(V)=q_{1:T}(\mathbf z^{\mathrm a})+\mathbf D r(\mathbf z^{\mathrm c}),
+ \qquad
+ \mathcal N_T=\operatorname{im}\Cov(\mathbf D r(\mathbf z^{\mathrm c})),
+\]
+where $\mathbf D$ is fixed, $q_{1:T}$ stacks the framewise action responses
+$q_t$, and $\mathcal N_T$ is the subspace of appearance variation.
+
+\begin{corollary}[Identification from additive features]
+\label{cor:additive-anchor-learning}
+Assume Condition~\ref{cond:anchor-identification}(i)--(ii), continuous
+$\Phi,q_t$, and finite covariance of $r(\mathbf z^{\mathrm c})$.
+Suppose $q_{1:T}(\mathbf a)-q_{1:T}(\mathbf a')\notin\mathcal N_T$
+for any distinct $\mathbf a,\mathbf a'\in\mathcal A_T$.
+Over the population model class above, with
+$d_{\mathrm a}\geq\operatorname{rank}\Cov(\Phi(V))-\dim\mathcal N_T$,
+every global minimizer of Eq.~\ref{eq:anchor-objective} with
+$\varphi(V^+)$ replaced by $\Phi(V^+)$ and squared Euclidean prediction
+error identifies the action block.
+\end{corollary}
+
+Averaging over appearance leaves an injective function of the action
+trajectory, allowing ordinary feature prediction to identify the action
+block (proof in Appendix~\ref{subsec:anchor-proof-identification}).
diff --git a/projects/video-action-editing/paper/source/3_method.tex b/projects/video-action-editing/paper/source/3_method.tex
new file mode 100644
index 0000000..6351798
--- /dev/null
+++ b/projects/video-action-editing/paper/source/3_method.tex
@@ -0,0 +1,135 @@
+\section{Method}
+\label{sec:method}
+VADD learns action representations from demonstrations and trains an
+editor to execute them under source conditioning
+(Figure~\ref{fig:vadd-method}). The editor receives source $V_{\mathrm s}$,
+instruction $i$, and demonstration $V_{\mathrm r}$, any video that performs
+the requested action, such as a reference recording or a text-to-video
+sample; it need not share the source's appearance or frame alignment.
+Section~\ref{subsec:experimental-setup} describes the demonstrations used
+in training and at inference.
+
+\begin{figure}[!t]
+ \centering
+ \includegraphics[width=\linewidth]{vadd_method_overview_anchor.pdf}
+ \caption{VADD overview.
+ (a) A temporal encoder extracts action anchors from full dynamic observations.
+ Normalized anchors (orange) and compressed V-JEPA2 features (teal) form
+ $\mathbf K$. Training uses same-video patch views for consistency,
+ dynamic-observation reconstruction, and KL matching to a history-conditioned
+ factorized prior.
+ (b) Source context, generated states, and phase-aligned action conditions
+ drive residual adapters during early sampling steps.
+ Clean predictions receive source-preservation and ordered behavior
+ supervision, with source, reversed, and shuffled videos as controls.
+ Dashed paths denote training; repeated $\widehat V_{0\mid k}$ ports refer
+ to the same prediction. Source/text preprocessing and the sampling loop
+ are condensed. Videos, tokens, and distributions are schematic.}
+ \label{fig:vadd-method}
+\end{figure}
+
+\subsection{Learning action representations from demonstrations}
+\label{subsec:method-anchor}
+
+Frozen backbone features of a demonstration yield dynamic observations $\mathbf O(V_{\mathrm r})\in\R^{L\times d_o}$ over $L$
+resampled temporal phases. Temporal centering removes constant offsets;
+signed changes and local change magnitudes retain direction and variation.
+Alternating spatial patches of the same video provide two training views.
+
+A bidirectional recurrent encoder produces a Gaussian posterior
+$q_\phi(\widetilde{\mathbf z}^{\mathrm a}_{1:L}\mid \mathbf O)$. Its mean sequence $\widehat{\mathbf z}^{\mathrm a}=E_{\mathrm a}(V_{\mathrm r})=\boldsymbol\mu_\phi(\mathbf O(V_{\mathrm r}))$ serves as the
+learned action anchor, with $d_{\mathrm a}=Ld_z$ after vectorization. A history-conditioned prior factorizes as
+\begin{equation}
+ p_\psi(\widetilde{\mathbf z}^{\mathrm a}_{1:L})
+ =\prod_{t=1}^{L}\prod_{j=1}^{d_z}
+ p_{\psi,t,j}(\widetilde z^{\mathrm a}_{t,j}\mid \widetilde{\mathbf z}^{\mathrm a}_{Synthetic source pairing · VADD← Video Action Editing
# Synthetic qualitative figure: two examples, four frames, three rows
+
+The composition follows the horizontal frame strips in `figures/intro.pdf`.
+Each example has Source, Reference, and ours rows. In the manually edited
+`my_layout` version, the editing instructions appear above the Source row.
+Original frame indices are 0, 27,
+53, and 80. All PNG assets retain the supplied 352 x 192 crop and exact bytes.
+
+## Versions
+
+- `synthetic_qualitative_my_layout.drawio` / `.pdf`: uses exactly the Source,
+ Reference, and ours images in the two user-selected `my` directories. The
+ instructions are **Push forward → Push right** and
+ **Push forward-left → Push back-right**, matching the displayed motions.
+ The user's manual layout is preserved. Six editable perspective arrows mark
+ the initial block motion in the first column of both examples (all three rows).
+ Each arrow has a white top face, blue-gray extruded sides, and a floor shadow;
+ its ground-plane projection narrows with distance. This is an illustrative
+ perspective treatment, not a calibrated scene reconstruction.
+ Their directions follow the visible displacement from frame 0 to frame 27;
+ the right Reference and Ours initially move up-right before their later turn.
+ Arrow positions and directions are recorded in `initial_motion_annotations.json`.
+ Per the author, these Source images are the actual inputs; see the
+ provenance discussion below. This version is the manuscript figure
+ (`figures/synthetic_qualitative.pdf`).
+- `synthetic_qualitative_paired.drawio` / `.pdf`: restores the archived Source
+ images for both experiments. All reference and output pixels remain the
+ same. Instructions are **Push back-right → Push right** and
+ **Push back-right → Push back-right**. The second example is a same-action
+ reference control, not a direction-changing edit.
+- `*_preview.png`: previews rendered from the exported PDFs.
+- `assets_manifest.json`: exact supplied paths, frame indices, hashes, and
+ provenance matches. All imported assets are retained under `assets/`.
+- `build_figure.py` and `*_scene.json`: generator and geometry snapshots of the
+ original proposals, before manual layout changes and motion annotations.
+
+Every image, text label, instruction, and divider is a separate native draw.io
+object. Images are embedded, so each `.drawio` file is self-contained.
+The PDF is exported directly from its corresponding draw.io document.
+Each perspective arrow is one native custom-stencil object that can be moved
+or resized independently. Its faces remain vector paths in the PDF; no frame
+pixels are changed. `style_motion_arrows.py` restyles only the six named arrows
+using their existing positions and directions, without rebuilding the layout.
+
+## Source/output pairing
+
+The left `my/source` matches C218 `train_family000_bearing2_source1` (forward).
+The right `my/source` matches C218 `train_family000_bearing3_source1`
+(forward-left). Meanwhile, the supplied `my/ours` assets match the existing
+C399 and C523 learned-edit frames byte-for-byte. The archived experiment
+folders record the back-right clip as Source. On 2026-09-26 the author
+confirmed that this archived Source record is wrong and that the `my/source`
+clips are the correct inputs, so the `my_layout` pairing is used in the
+manuscript. The `paired` version is kept only as a record of the archived
+folders.
+
+The output row contains the supplied historical C399/C523 synthetic
+learned-consumer outputs. Figure editing performs
+layout and byte-preserving asset import only; it does not run a new experiment
+or establish a new full-method, convergence, generalization, or success claim.
+`expected` files in the supplied directories are deliberately not used as ours.
+
+The `my_layout` drawing is exported by the author to
+`figures/synthetic_qualitative.pdf` and included in Section 5
+(`fig:synthetic-qualitative`). The Overleaf export manifest is unchanged.
+
+## Export the current drawings
+
+From the repository root:
+
+```sh
+/Applications/draw.io.app/Contents/MacOS/draw.io --export --format pdf --crop --output working/figures/synthetic_qualitative_v1/synthetic_qualitative_my_layout.pdf working/figures/synthetic_qualitative_v1/synthetic_qualitative_my_layout.drawio
+/Applications/draw.io.app/Contents/MacOS/draw.io --export --format pdf --crop --output working/figures/synthetic_qualitative_v1/synthetic_qualitative_paired.pdf working/figures/synthetic_qualitative_v1/synthetic_qualitative_paired.drawio
+pdftoppm -scale-to 2600 -png -singlefile working/figures/synthetic_qualitative_v1/synthetic_qualitative_my_layout.pdf working/figures/synthetic_qualitative_v1/synthetic_qualitative_my_layout_preview
+pdftoppm -scale-to 2600 -png -singlefile working/figures/synthetic_qualitative_v1/synthetic_qualitative_paired.pdf working/figures/synthetic_qualitative_v1/synthetic_qualitative_paired_preview
+```
+
+`build_figure.py` regenerates the original proposals and overwrites the current
+draw.io files, including manual edits and motion arrows. Use the export commands
+above to refresh PDFs and previews from the edited drawings. If deliberately
+recreating the proposals, save the edited drawings separately first. The
+generator's `--import-assets` option imports the original supplied files and
+rejects an existing asset with different bytes.
+
\ No newline at end of file
diff --git a/projects/video-action-editing/paper/vadd_method_overview_anchor.pdf b/projects/video-action-editing/paper/vadd_method_overview_anchor.pdf
new file mode 100644
index 0000000..e3edd87
Binary files /dev/null and b/projects/video-action-editing/paper/vadd_method_overview_anchor.pdf differ
diff --git a/projects/video-action-editing/paper/vadd_method_overview_anchor.png b/projects/video-action-editing/paper/vadd_method_overview_anchor.png
new file mode 100644
index 0000000..e6a7283
Binary files /dev/null and b/projects/video-action-editing/paper/vadd_method_overview_anchor.png differ
diff --git a/projects/video-action-editing/qualitative/decode-validation.json b/projects/video-action-editing/qualitative/decode-validation.json
new file mode 100644
index 0000000..e20b764
--- /dev/null
+++ b/projects/video-action-editing/qualitative/decode-validation.json
@@ -0,0 +1,617 @@
+[
+ {
+ "path": "qualitative/omni/media/Te_Camera_speed_change_04/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Camera_speed_change_04/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Camera_speed_change_04/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Camera_speed_change_04/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Camera_speed_change_04/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Camera_speed_change_04/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Camera_speed_change_04/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Camera_speed_change_04/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Camera_speed_change_04/zebra.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Facial_expression_change_06/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Facial_expression_change_06/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Facial_expression_change_06/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Facial_expression_change_06/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Facial_expression_change_06/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Facial_expression_change_06/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Facial_expression_change_06/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Facial_expression_change_06/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Facial_expression_change_06/zebra.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_deletion_01/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_deletion_01/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_deletion_01/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_deletion_01/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_deletion_01/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_deletion_01/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_deletion_01/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_deletion_01/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_deletion_01/zebra.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_02/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_02/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_02/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_02/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_02/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_02/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_02/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_02/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_02/zebra.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_06/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_06/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_06/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_06/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_06/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_06/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_06/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_06/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/omni/media/Te_Motion_frequency_change_06/zebra.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/15_mev_d045859258943df0/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/15_mev_d045859258943df0/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/15_mev_d045859258943df0/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/15_mev_d045859258943df0/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/15_mev_d045859258943df0/mev160_fm_only_u160.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/15_mev_d045859258943df0/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/15_mev_d045859258943df0/real_target.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/15_mev_d045859258943df0/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/15_mev_d045859258943df0/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/15_mev_d045859258943df0/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/20_mev_059fef481bb29b38/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/20_mev_059fef481bb29b38/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/20_mev_059fef481bb29b38/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/20_mev_059fef481bb29b38/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/20_mev_059fef481bb29b38/mev160_fm_only_u160.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/20_mev_059fef481bb29b38/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/20_mev_059fef481bb29b38/real_target.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/20_mev_059fef481bb29b38/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/20_mev_059fef481bb29b38/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/20_mev_059fef481bb29b38/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/32_mev_a35540d41c3d7989/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/32_mev_a35540d41c3d7989/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/32_mev_a35540d41c3d7989/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/32_mev_a35540d41c3d7989/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/32_mev_a35540d41c3d7989/mev160_fm_only_u160.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/32_mev_a35540d41c3d7989/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/32_mev_a35540d41c3d7989/real_target.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/32_mev_a35540d41c3d7989/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/32_mev_a35540d41c3d7989/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/32_mev_a35540d41c3d7989/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/34_mev_f308ebe4cd1b3b79/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/34_mev_f308ebe4cd1b3b79/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/34_mev_f308ebe4cd1b3b79/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/34_mev_f308ebe4cd1b3b79/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/34_mev_f308ebe4cd1b3b79/mev160_fm_only_u160.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/34_mev_f308ebe4cd1b3b79/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/34_mev_f308ebe4cd1b3b79/real_target.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/34_mev_f308ebe4cd1b3b79/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/34_mev_f308ebe4cd1b3b79/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/34_mev_f308ebe4cd1b3b79/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/35_mev_1e9a413e9af09b7d/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/35_mev_1e9a413e9af09b7d/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/35_mev_1e9a413e9af09b7d/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/35_mev_1e9a413e9af09b7d/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/35_mev_1e9a413e9af09b7d/mev160_fm_only_u160.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/35_mev_1e9a413e9af09b7d/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/35_mev_1e9a413e9af09b7d/real_target.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/35_mev_1e9a413e9af09b7d/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/35_mev_1e9a413e9af09b7d/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/35_mev_1e9a413e9af09b7d/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/39_mev_c314f469fbaeb749/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/39_mev_c314f469fbaeb749/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/39_mev_c314f469fbaeb749/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/39_mev_c314f469fbaeb749/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/39_mev_c314f469fbaeb749/mev160_fm_only_u160.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/39_mev_c314f469fbaeb749/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/39_mev_c314f469fbaeb749/real_target.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/39_mev_c314f469fbaeb749/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/39_mev_c314f469fbaeb749/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/39_mev_c314f469fbaeb749/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/delete_action/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/delete_action/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/delete_action/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/delete_action/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/delete_action/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/delete_action/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/delete_action/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/delete_action/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/delete_action/zebra.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/man_laugh/frozen_bernini.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/man_laugh/kiwi_edit_5b_instruct_only.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/man_laugh/liveedit_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/man_laugh/lucy_edit_1p1_dev_5b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/man_laugh/omnivideo2_official_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/man_laugh/reco_ori_wan21_1p3b.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/man_laugh/source.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/man_laugh/vace_wan21_1p3b_fullframe.mp4",
+ "ok": true,
+ "errors": ""
+ },
+ {
+ "path": "qualitative/real/man_laugh/zebra.mp4",
+ "ok": true,
+ "errors": ""
+ }
+]
diff --git a/projects/video-action-editing/qualitative/manifest.json b/projects/video-action-editing/qualitative/manifest.json
new file mode 100644
index 0000000..8af4697
--- /dev/null
+++ b/projects/video-action-editing/qualitative/manifest.json
@@ -0,0 +1,962 @@
+[
+ {
+ "asset": "../qualitative/omni/selection.json",
+ "sha256": "84cd6009ae3da63f710f32de0c33997a574e1ea202f35e8371b20b87094c6acb",
+ "bytes": 51370
+ },
+ {
+ "asset": "../qualitative/omni/ranking_all200.json",
+ "sha256": "b03bde3c2e8673cad012e922ccff0049f2385fd4d382ebb6e7b1bdc8d9b66243",
+ "bytes": 34386
+ },
+ {
+ "asset": "../qualitative/omni/download_audit.json",
+ "sha256": "27b80926718aef62697ec216ebb3201f0e692bda95281ab2246485913413fe3e",
+ "bytes": 20183
+ },
+ {
+ "asset": "../qualitative/omni/media_validation.json",
+ "sha256": "c956781e90b8360963d6c9e9a643d8368ad4fa47d97867a8d53480fdfc60aa59",
+ "bytes": 18409
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_02/source.mp4",
+ "sha256": "eee3a86a3441fb2617dcc8323c157d39fc83dc2bd2c0edf1ea860cd6d9c903d1",
+ "bytes": 938527
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_02/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "9fc6f9989ff621d63f3288406708eed133a514e7a5f5668eea57832e82ada495",
+ "bytes": 529899
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_02/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "7faa97ec69c6559f1251d072e506a877e8e85f52239893bebd3a33d66b6269db",
+ "bytes": 749001
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_02/reco_ori_wan21_1p3b.mp4",
+ "sha256": "90db4129783fa802f016a01e0bd78b02327445d4a423c3fb499b103d2b75c21d",
+ "bytes": 854641
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_02/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "4fc6bd5fdcedb2703a879b09cc0460dc9da0646ab8d9774fa4ba828f46a4e972",
+ "bytes": 1345507
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_02/zebra.mp4",
+ "sha256": "20c872bafe12b8de1073240051242732936beb02b3a2ea716fa18a5194ff70de",
+ "bytes": 831981
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_02/liveedit_wan21_1p3b.mp4",
+ "sha256": "a37e2101b84e691a17865a185c5900287036f1443d241d929a0fbe8c65180080",
+ "bytes": 967998
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_02/frozen_bernini.mp4",
+ "sha256": "e166a44a9312219d78537f13f49e8e04800a1814f02290ea39334ff7efddafd9",
+ "bytes": 881279
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_02/omnivideo2_official_1p3b.mp4",
+ "sha256": "d030b29c60403f5c51b9c8d9272eb81853e270f81199ec05918b1d9db971dc78",
+ "bytes": 1182261
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Facial_expression_change_06/source.mp4",
+ "sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "bytes": 1655362
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Facial_expression_change_06/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "3dbb96889496be16482f84a759a34a4e38e39dbde4a376b4424054efd06ea740",
+ "bytes": 1085868
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Facial_expression_change_06/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "f0285123defe401fe6d9b707716de369dc8162402f64a59748a756752dfb0e2b",
+ "bytes": 825523
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Facial_expression_change_06/reco_ori_wan21_1p3b.mp4",
+ "sha256": "a23c0cb3676dc3904957c80dad444a59651862972ecc31155bc01fd40d525246",
+ "bytes": 865431
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Facial_expression_change_06/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "f2194f1c764e6f1f00b1e9e59c0c213e4e574dfa3024cef0d565ba165d081748",
+ "bytes": 1597654
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Facial_expression_change_06/zebra.mp4",
+ "sha256": "231b466565101154bf9185a812dbffa1195988d8873b746f82d393c913ad66a1",
+ "bytes": 1491176
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Facial_expression_change_06/liveedit_wan21_1p3b.mp4",
+ "sha256": "8ba93c48945cd80e3248682620976249e165313f5042c4a7ec04166ee8a28ddf",
+ "bytes": 2023071
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Facial_expression_change_06/frozen_bernini.mp4",
+ "sha256": "2234842811d1243ecb892b5beeb78ce3ef26fc0222a9f92cf067e59fede4fd2c",
+ "bytes": 1696848
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Facial_expression_change_06/omnivideo2_official_1p3b.mp4",
+ "sha256": "95271b71d0cc214d83de53f2946b89778e77e076e26b4bc6a715372cfd5e1fa7",
+ "bytes": 3021981
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_06/source.mp4",
+ "sha256": "89033eb8d081cf564f28ff41111f25fcd6bd7f41604c0f4d3e54bb6f950f5ee8",
+ "bytes": 3961122
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_06/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "69143741503ec5c260fb5682fba38a9ac0a8d95a23689f582828454821c4c746",
+ "bytes": 1506545
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_06/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "8cdbcf5381196d017811d9f74464977bc946bb670dd9fb07d60b46b95278ee31",
+ "bytes": 1926421
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_06/reco_ori_wan21_1p3b.mp4",
+ "sha256": "e59014ce6f558d6fb175cac92be6a90e8af459a6bc05eb255878ce42821a007b",
+ "bytes": 2486265
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_06/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "d54f0a371d7519fd205d5fb20f38520953477a5d6bd61d807e58d735861f4c4f",
+ "bytes": 3857817
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_06/zebra.mp4",
+ "sha256": "99e7098bf8ca8f327d9acf1c72b3c88b82ca7e79c3b7fc22c0fd503f53a88f23",
+ "bytes": 1732969
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_06/liveedit_wan21_1p3b.mp4",
+ "sha256": "920905582914e55b249fb889a697a64f987e527e4583fb70f912d6ce4d664164",
+ "bytes": 2745445
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_06/frozen_bernini.mp4",
+ "sha256": "197736d1ded8d96cfe65d3fd3a8982b9866a60a19d6b5799bfd337258db95700",
+ "bytes": 1596752
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_frequency_change_06/omnivideo2_official_1p3b.mp4",
+ "sha256": "0ec0232e51cbdbced06ee86e9c81712facbe845ba0689521e2f45547925dd786",
+ "bytes": 3388422
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_deletion_01/source.mp4",
+ "sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "bytes": 1037462
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_deletion_01/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "365a6e85a43a125865a6c94e892c32490234771de2cea0874c4ba9a7db803d50",
+ "bytes": 466006
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_deletion_01/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "bacf48ee36c1b5fad294c9ea3a3f6376792946c23297800b67f197ab3bd1d0be",
+ "bytes": 597859
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_deletion_01/reco_ori_wan21_1p3b.mp4",
+ "sha256": "8febc79000879f423a5236cff8d712953999a152911f8d606c50bff0d458920f",
+ "bytes": 958841
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_deletion_01/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "c4a5537e65793a14c7f0aa816056105672f6864ff8b4bf89230f08a1ea28e371",
+ "bytes": 1414568
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_deletion_01/zebra.mp4",
+ "sha256": "1d4de93dce10739e46069ec9f269a92de765c2f970dfa85d8de42a1875640ddd",
+ "bytes": 584758
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_deletion_01/liveedit_wan21_1p3b.mp4",
+ "sha256": "b6758782a0ed97a2aae04121b61ad61fe69b481bae4a9220f4be2aeae381a889",
+ "bytes": 1194297
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_deletion_01/frozen_bernini.mp4",
+ "sha256": "cd28f95da224595c3b323af7042fe788a08cb033637e3b958144287dd92a241e",
+ "bytes": 645612
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Motion_deletion_01/omnivideo2_official_1p3b.mp4",
+ "sha256": "362e2f0a8b72ccdac9fc6bb89eca33b5f54333d4b5454e7ba8c6ecf249aaa699",
+ "bytes": 1071023
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Camera_speed_change_04/source.mp4",
+ "sha256": "0fddc8d23c7f6f5d55658de1ee5de4d588c20a78b6acf0344a7f467a5f6e6b84",
+ "bytes": 1393842
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Camera_speed_change_04/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "a47894d39923fd8907bf97af4b40b1b24bcb4b2bccf4090e11ea99ea2972faa9",
+ "bytes": 1044453
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Camera_speed_change_04/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "5e4e29619c5a16d24b9f9caf1459a9f9295d34669588d839f2b9ad6bc39627df",
+ "bytes": 692403
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Camera_speed_change_04/reco_ori_wan21_1p3b.mp4",
+ "sha256": "765f0c272a9f489fce69bdfa46a1d3826ad52af4832e031eb80c1a9f8696bb9e",
+ "bytes": 1865906
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Camera_speed_change_04/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "4f1a179eb4ae4737ffe1e7555b0ddd76413895f53090bc333900cf619a51452b",
+ "bytes": 1816057
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Camera_speed_change_04/zebra.mp4",
+ "sha256": "72a872773f4073862d9b67051d135f9f587bce0c2ca844fe4d918c8ed5e8c8de",
+ "bytes": 1147466
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Camera_speed_change_04/liveedit_wan21_1p3b.mp4",
+ "sha256": "6c92a2c7edc0ba610fcf39579754239b2996cd457be2d8133e311c26d5e9b451",
+ "bytes": 1790100
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Camera_speed_change_04/frozen_bernini.mp4",
+ "sha256": "4da6ebd8b81e975b4cdf8f7532c3c36b84dc792958e72c4bdb3c77653c7a3392",
+ "bytes": 1692502
+ },
+ {
+ "asset": "../qualitative/omni/media/Te_Camera_speed_change_04/omnivideo2_official_1p3b.mp4",
+ "sha256": "7e909aa8b0a5bd039c10999b362a1f5de10926de3be2e8a4859e8b60d63a349b",
+ "bytes": 2969904
+ },
+ {
+ "asset": "../qualitative/real/15_mev_d045859258943df0/source.mp4",
+ "sha256": "bcf47f742b054db8cefaa3b7ad3d556f9bcbbe488fc7193bac1164262fe66ace",
+ "bytes": 1022213
+ },
+ {
+ "asset": "../qualitative/real/15_mev_d045859258943df0/mev160_fm_only_u160.mp4",
+ "sha256": "b86868dbc2218a6169b80edec35d8bc0c2698d33d4841f54c9c211e317d17b23",
+ "bytes": 2875407
+ },
+ {
+ "asset": "../qualitative/real/15_mev_d045859258943df0/frozen_bernini.mp4",
+ "sha256": "d97b74ac16c05f82ae1568e3d378e6e57fcc89c173dbb668ceba4157dbb119c8",
+ "bytes": 3393629
+ },
+ {
+ "asset": "../qualitative/real/15_mev_d045859258943df0/real_target.mp4",
+ "sha256": "fde32c88e8a723828a382aea5272ae41ca0e4f8f9e941254cd0cf496dd16833d",
+ "bytes": 3587767
+ },
+ {
+ "asset": "../qualitative/real/15_mev_d045859258943df0/omnivideo2_official_1p3b.mp4",
+ "sha256": "e3329f0089023749727ecb5108eb5c68193a556c169a8075c72d79ffac881302",
+ "bytes": 5923006
+ },
+ {
+ "asset": "../qualitative/real/15_mev_d045859258943df0/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "e94850b835b97e76d108710fced2160eecef71577d1887477dc1968c2d990d56",
+ "bytes": 275673
+ },
+ {
+ "asset": "../qualitative/real/15_mev_d045859258943df0/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "b3e40ae539d10ad1652f62fd02447b2cb5d7915ec3b1f40a60f974dc518c01f4",
+ "bytes": 4918050
+ },
+ {
+ "asset": "../qualitative/real/15_mev_d045859258943df0/liveedit_wan21_1p3b.mp4",
+ "sha256": "923782590f737fc0dd561af1af4065a969da358417d0305a77b7386fa06e3713",
+ "bytes": 7848212
+ },
+ {
+ "asset": "../qualitative/real/15_mev_d045859258943df0/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "f02bab775e9d5fb42ea215f0b41b51b2518068bc9df3b8b21f0fc93e28c72fd7",
+ "bytes": 190151
+ },
+ {
+ "asset": "../qualitative/real/15_mev_d045859258943df0/reco_ori_wan21_1p3b.mp4",
+ "sha256": "0bcac2bf2f1b66a7b6b710683c5d6a4b2e5eaf0e77eb23da1594862c433e5e20",
+ "bytes": 7411778
+ },
+ {
+ "asset": "../qualitative/real/20_mev_059fef481bb29b38/source.mp4",
+ "sha256": "46c5579ec129e6c834ed54c6089db579d42d527b218d93470c41955090680cec",
+ "bytes": 667296
+ },
+ {
+ "asset": "../qualitative/real/20_mev_059fef481bb29b38/mev160_fm_only_u160.mp4",
+ "sha256": "217a19226830187a58a989badeeb7d3afabebec80b0cfe412cf3f25cbfde2668",
+ "bytes": 4134554
+ },
+ {
+ "asset": "../qualitative/real/20_mev_059fef481bb29b38/frozen_bernini.mp4",
+ "sha256": "9652591271773444afd470983f83e93baa1632e90484281943f00f7889b0402b",
+ "bytes": 3440136
+ },
+ {
+ "asset": "../qualitative/real/20_mev_059fef481bb29b38/real_target.mp4",
+ "sha256": "781a12123838eb842730223b998f6dce5bce28b4cfe7cd124aa93cc5f78acb1f",
+ "bytes": 7682512
+ },
+ {
+ "asset": "../qualitative/real/20_mev_059fef481bb29b38/omnivideo2_official_1p3b.mp4",
+ "sha256": "ffcd8a1c0514c2606cac5b133c80fed45041b5c2ab2acdc385eb3b477c403b20",
+ "bytes": 4336422
+ },
+ {
+ "asset": "../qualitative/real/20_mev_059fef481bb29b38/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "8bdd24f357ab6af246fff855bfe7769d7c1ac5902f451e0ec21e3b6d486f3715",
+ "bytes": 223880
+ },
+ {
+ "asset": "../qualitative/real/20_mev_059fef481bb29b38/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "4050c1bcd1f1e535996ff9dfb5920c214bc2719b0233597baa86823fb7266e59",
+ "bytes": 4755622
+ },
+ {
+ "asset": "../qualitative/real/20_mev_059fef481bb29b38/liveedit_wan21_1p3b.mp4",
+ "sha256": "a306537ead472f19c6b5c182c39b6b20bbd663efeff0cbe2e299f20c8f07407f",
+ "bytes": 8027236
+ },
+ {
+ "asset": "../qualitative/real/20_mev_059fef481bb29b38/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "7de4c4ca05916c0718c3431c3c617724194d846cfe699849249e7e275dd3c3b3",
+ "bytes": 321819
+ },
+ {
+ "asset": "../qualitative/real/20_mev_059fef481bb29b38/reco_ori_wan21_1p3b.mp4",
+ "sha256": "53970bd97c770bb50cb60be768807e7eec27355357c16bf82d022802a4eb68f8",
+ "bytes": 5579035
+ },
+ {
+ "asset": "../qualitative/real/32_mev_a35540d41c3d7989/source.mp4",
+ "sha256": "ecb6d1d0653b24288355793a9c9d0f1507a3efa41d46c5900824786815293272",
+ "bytes": 589143
+ },
+ {
+ "asset": "../qualitative/real/32_mev_a35540d41c3d7989/mev160_fm_only_u160.mp4",
+ "sha256": "eae3b9faf2488a9644cc5dac6f9b20476548dbed04fd191845cb74249fa4935b",
+ "bytes": 1474881
+ },
+ {
+ "asset": "../qualitative/real/32_mev_a35540d41c3d7989/frozen_bernini.mp4",
+ "sha256": "0474a8bff31c35e9a798f15f2856279c3018081f4ddc8fce5d0af33251a43ed3",
+ "bytes": 2209548
+ },
+ {
+ "asset": "../qualitative/real/32_mev_a35540d41c3d7989/real_target.mp4",
+ "sha256": "94cbfd35fa002eaf62efb9f77fcbf2b7e686320daccf4d89487d5f25756d7c0b",
+ "bytes": 953460
+ },
+ {
+ "asset": "../qualitative/real/32_mev_a35540d41c3d7989/omnivideo2_official_1p3b.mp4",
+ "sha256": "7bf26411385d098d5b658562d9352a73501d382c1afce0cc0a2d2cc28276dc24",
+ "bytes": 1896249
+ },
+ {
+ "asset": "../qualitative/real/32_mev_a35540d41c3d7989/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "b4f24c17daa1c85efabe838c98fd333622c89ad6b69b716dee6ed52a23b4ccc8",
+ "bytes": 235629
+ },
+ {
+ "asset": "../qualitative/real/32_mev_a35540d41c3d7989/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "7adc65576001ba8cac78a4580c98053b2894cd5ab53a200c328314374700beca",
+ "bytes": 2339605
+ },
+ {
+ "asset": "../qualitative/real/32_mev_a35540d41c3d7989/liveedit_wan21_1p3b.mp4",
+ "sha256": "ad06da1f58f1f7e07c2d413d84ef97310b6e6de71ce30b553882929f13416112",
+ "bytes": 4108275
+ },
+ {
+ "asset": "../qualitative/real/32_mev_a35540d41c3d7989/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "ba9932a92c98b15e3fe086b8db697bb38cc883dbe07759543272b1d330d6e66c",
+ "bytes": 274181
+ },
+ {
+ "asset": "../qualitative/real/32_mev_a35540d41c3d7989/reco_ori_wan21_1p3b.mp4",
+ "sha256": "207f2269fc3880368d7155ecb960634a3c3d6352d602411445135da990ae4dc2",
+ "bytes": 3981733
+ },
+ {
+ "asset": "../qualitative/real/34_mev_f308ebe4cd1b3b79/source.mp4",
+ "sha256": "913c4764ad79ef26f69f7e10d3693718313f1206cdb2f21ab5a668f7966136da",
+ "bytes": 719461
+ },
+ {
+ "asset": "../qualitative/real/34_mev_f308ebe4cd1b3b79/mev160_fm_only_u160.mp4",
+ "sha256": "71290f13a4acfe1452b7def568fd6076a204603e36d2f14bce0d956cc32d5a1f",
+ "bytes": 2068859
+ },
+ {
+ "asset": "../qualitative/real/34_mev_f308ebe4cd1b3b79/frozen_bernini.mp4",
+ "sha256": "d7d511f4c8bc7e7298e44cf911ef1495e6a70215c4f0ef448b6238b3d3e0773b",
+ "bytes": 2205091
+ },
+ {
+ "asset": "../qualitative/real/34_mev_f308ebe4cd1b3b79/real_target.mp4",
+ "sha256": "2d03f489132a86aeb5bb8c0fa383c709366659a06bc73d4015b139a634aadc32",
+ "bytes": 2867360
+ },
+ {
+ "asset": "../qualitative/real/34_mev_f308ebe4cd1b3b79/omnivideo2_official_1p3b.mp4",
+ "sha256": "3900233d518752723f8c7037fa00fcbf89836175ecdfe8f26c687c5f9f707ca5",
+ "bytes": 2473141
+ },
+ {
+ "asset": "../qualitative/real/34_mev_f308ebe4cd1b3b79/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "045949bbe07248e31aa0bb634a237f0f99d9a179a86bb23a4771ae7823f219ba",
+ "bytes": 262366
+ },
+ {
+ "asset": "../qualitative/real/34_mev_f308ebe4cd1b3b79/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "f6be33e019392a33d8577375f68926641bd7f0fdbf33353fd91083a47500e45c",
+ "bytes": 2993485
+ },
+ {
+ "asset": "../qualitative/real/34_mev_f308ebe4cd1b3b79/liveedit_wan21_1p3b.mp4",
+ "sha256": "3b7b46fa4105400d2be3c22fb25c93725d95201a93c8ff6844b41faa9ae688a1",
+ "bytes": 5188827
+ },
+ {
+ "asset": "../qualitative/real/34_mev_f308ebe4cd1b3b79/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "89f95c44bd87656b3c4acd601e150b3e5dba6f33f4760dc05a01df4b2673eceb",
+ "bytes": 215565
+ },
+ {
+ "asset": "../qualitative/real/34_mev_f308ebe4cd1b3b79/reco_ori_wan21_1p3b.mp4",
+ "sha256": "51678c7dbe2e116616aa0b495ca9649a80b6ec0f95f755129f461e29dfefed8e",
+ "bytes": 5273764
+ },
+ {
+ "asset": "../qualitative/real/35_mev_1e9a413e9af09b7d/source.mp4",
+ "sha256": "813a16c21affab251a596bd44884eb305be15348206e8fa909f0ca1bc703d75d",
+ "bytes": 193882
+ },
+ {
+ "asset": "../qualitative/real/35_mev_1e9a413e9af09b7d/mev160_fm_only_u160.mp4",
+ "sha256": "6da62956af13614338557beea76f06ba431065a7b3cc35f70b0301f179fe0bb5",
+ "bytes": 1603703
+ },
+ {
+ "asset": "../qualitative/real/35_mev_1e9a413e9af09b7d/frozen_bernini.mp4",
+ "sha256": "5bdd1c7ef98cb1617ac0d67f8cae2a1443bcf4cf2d04d806ee5fda5663af5f91",
+ "bytes": 1859473
+ },
+ {
+ "asset": "../qualitative/real/35_mev_1e9a413e9af09b7d/real_target.mp4",
+ "sha256": "66ee70f2f4d85c5b831e1519c294c8a215be5dec383bfaab2ac0b6916c4f0e45",
+ "bytes": 3899269
+ },
+ {
+ "asset": "../qualitative/real/35_mev_1e9a413e9af09b7d/omnivideo2_official_1p3b.mp4",
+ "sha256": "8acb1bb33f9784faf0541b4e2efe9f051b1acfc67d9e4d751c91cef3551d18ab",
+ "bytes": 890992
+ },
+ {
+ "asset": "../qualitative/real/35_mev_1e9a413e9af09b7d/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "3672364f5a81618ef5fb3468973d7d8443671fc2a762aaee00e34486f0f34278",
+ "bytes": 93396
+ },
+ {
+ "asset": "../qualitative/real/35_mev_1e9a413e9af09b7d/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "6b0367e01cc0f6d27fa0ea75cd6a582079d49bec565900cafc9eb9fb2c5b3e30",
+ "bytes": 1494699
+ },
+ {
+ "asset": "../qualitative/real/35_mev_1e9a413e9af09b7d/liveedit_wan21_1p3b.mp4",
+ "sha256": "b689d8b8afde74002e30f06c830db1976de033590305d933e07d56c33d7b4654",
+ "bytes": 3738415
+ },
+ {
+ "asset": "../qualitative/real/35_mev_1e9a413e9af09b7d/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "591f59bdb70c38d159c89aa3d01652217322e740ecd31333fe515eb056ea18a7",
+ "bytes": 82058
+ },
+ {
+ "asset": "../qualitative/real/35_mev_1e9a413e9af09b7d/reco_ori_wan21_1p3b.mp4",
+ "sha256": "cca0bb6599c9ce29ae3397069511d2468f59b9fb4d40364917fb983c1834dc6d",
+ "bytes": 3327418
+ },
+ {
+ "asset": "../qualitative/real/39_mev_c314f469fbaeb749/source.mp4",
+ "sha256": "faa139f727bb2e5c2da4a6af258e974d556b4549ab77b6d32426402eef897eba",
+ "bytes": 605944
+ },
+ {
+ "asset": "../qualitative/real/39_mev_c314f469fbaeb749/mev160_fm_only_u160.mp4",
+ "sha256": "49d3dca9214c99e26b6033ec04d780efbf14b06fa5fed08e5320a132ff2df4dd",
+ "bytes": 1483256
+ },
+ {
+ "asset": "../qualitative/real/39_mev_c314f469fbaeb749/frozen_bernini.mp4",
+ "sha256": "a82bd12dd0532f0499bb92061f931d8603e73b29c97b7e5a9c1facc46d996a9d",
+ "bytes": 2000563
+ },
+ {
+ "asset": "../qualitative/real/39_mev_c314f469fbaeb749/real_target.mp4",
+ "sha256": "b96e996d5c4e506c70047dd4e75330914acb2ea25e637b0939b000ed2dec74d8",
+ "bytes": 2175526
+ },
+ {
+ "asset": "../qualitative/real/39_mev_c314f469fbaeb749/omnivideo2_official_1p3b.mp4",
+ "sha256": "0e43bd888ac4485ae61f1e36f1b090f8853e0dc6f8d0b11ce5a111b206ff6fed",
+ "bytes": 2358400
+ },
+ {
+ "asset": "../qualitative/real/39_mev_c314f469fbaeb749/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "9d289c7d6b37655f42fc189465a44d3cdbb8f6c854efda9f960334d2d2011d01",
+ "bytes": 252278
+ },
+ {
+ "asset": "../qualitative/real/39_mev_c314f469fbaeb749/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "2432aa5ccb70173ae6757a55eb2fd27a7630b85b5208fc925b982de12762fd21",
+ "bytes": 2966435
+ },
+ {
+ "asset": "../qualitative/real/39_mev_c314f469fbaeb749/liveedit_wan21_1p3b.mp4",
+ "sha256": "697c86f5bc6b0bf507b71da223116b49a493a8669cd0bf91d9d3c68fd07f65ae",
+ "bytes": 5018690
+ },
+ {
+ "asset": "../qualitative/real/39_mev_c314f469fbaeb749/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "7e4eeba9c7ec8ff59c678714b26d319e885c667eee60b7c84b44073373293c25",
+ "bytes": 244143
+ },
+ {
+ "asset": "../qualitative/real/39_mev_c314f469fbaeb749/reco_ori_wan21_1p3b.mp4",
+ "sha256": "5e76c5851301e50689ff11d4c38007e6f256a5c64eb3b5590ff7f029efd35260",
+ "bytes": 4959413
+ },
+ {
+ "asset": "../qualitative/real/delete_action/source.mp4",
+ "sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "bytes": 1037462
+ },
+ {
+ "asset": "../qualitative/real/delete_action/zebra.mp4",
+ "sha256": "1d4de93dce10739e46069ec9f269a92de765c2f970dfa85d8de42a1875640ddd",
+ "bytes": 584758
+ },
+ {
+ "asset": "../qualitative/real/delete_action/frozen_bernini.mp4",
+ "sha256": "cd28f95da224595c3b323af7042fe788a08cb033637e3b958144287dd92a241e",
+ "bytes": 645612
+ },
+ {
+ "asset": "../qualitative/real/delete_action/omnivideo2_official_1p3b.mp4",
+ "sha256": "362e2f0a8b72ccdac9fc6bb89eca33b5f54333d4b5454e7ba8c6ecf249aaa699",
+ "bytes": 1071023
+ },
+ {
+ "asset": "../qualitative/real/delete_action/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "bacf48ee36c1b5fad294c9ea3a3f6376792946c23297800b67f197ab3bd1d0be",
+ "bytes": 597859
+ },
+ {
+ "asset": "../qualitative/real/delete_action/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "c4a5537e65793a14c7f0aa816056105672f6864ff8b4bf89230f08a1ea28e371",
+ "bytes": 1414568
+ },
+ {
+ "asset": "../qualitative/real/delete_action/liveedit_wan21_1p3b.mp4",
+ "sha256": "b6758782a0ed97a2aae04121b61ad61fe69b481bae4a9220f4be2aeae381a889",
+ "bytes": 1194297
+ },
+ {
+ "asset": "../qualitative/real/delete_action/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "365a6e85a43a125865a6c94e892c32490234771de2cea0874c4ba9a7db803d50",
+ "bytes": 466006
+ },
+ {
+ "asset": "../qualitative/real/delete_action/reco_ori_wan21_1p3b.mp4",
+ "sha256": "8febc79000879f423a5236cff8d712953999a152911f8d606c50bff0d458920f",
+ "bytes": 958841
+ },
+ {
+ "asset": "../qualitative/real/man_laugh/source.mp4",
+ "sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "bytes": 1655362
+ },
+ {
+ "asset": "../qualitative/real/man_laugh/zebra.mp4",
+ "sha256": "231b466565101154bf9185a812dbffa1195988d8873b746f82d393c913ad66a1",
+ "bytes": 1491176
+ },
+ {
+ "asset": "../qualitative/real/man_laugh/frozen_bernini.mp4",
+ "sha256": "2234842811d1243ecb892b5beeb78ce3ef26fc0222a9f92cf067e59fede4fd2c",
+ "bytes": 1696848
+ },
+ {
+ "asset": "../qualitative/real/man_laugh/omnivideo2_official_1p3b.mp4",
+ "sha256": "95271b71d0cc214d83de53f2946b89778e77e076e26b4bc6a715372cfd5e1fa7",
+ "bytes": 3021981
+ },
+ {
+ "asset": "../qualitative/real/man_laugh/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "f0285123defe401fe6d9b707716de369dc8162402f64a59748a756752dfb0e2b",
+ "bytes": 825523
+ },
+ {
+ "asset": "../qualitative/real/man_laugh/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "f2194f1c764e6f1f00b1e9e59c0c213e4e574dfa3024cef0d565ba165d081748",
+ "bytes": 1597654
+ },
+ {
+ "asset": "../qualitative/real/man_laugh/liveedit_wan21_1p3b.mp4",
+ "sha256": "8ba93c48945cd80e3248682620976249e165313f5042c4a7ec04166ee8a28ddf",
+ "bytes": 2023071
+ },
+ {
+ "asset": "../qualitative/real/man_laugh/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "3dbb96889496be16482f84a759a34a4e38e39dbde4a376b4424054efd06ea740",
+ "bytes": 1085868
+ },
+ {
+ "asset": "../qualitative/real/man_laugh/reco_ori_wan21_1p3b.mp4",
+ "sha256": "a23c0cb3676dc3904957c80dad444a59651862972ecc31155bc01fd40d525246",
+ "bytes": 865431
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/source/frame_000000_crop_352x192.png",
+ "sha256": "ed1df292d07dcc4a16c73a6b3e5941e21a5a141739fd0696ed03f4104ad4a0c4",
+ "bytes": 16894
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/source/frame_000027_crop_352x192.png",
+ "sha256": "825b35e3386160645675585638ceff2ec87f53088b0272f8983e01770cf56e6a",
+ "bytes": 16961
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/source/frame_000053_crop_352x192.png",
+ "sha256": "2bb48e9ea107491e0cb097d56e77fa1dba0e16cc9f9e0efcfe089f24177a9cb8",
+ "bytes": 15933
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/source/frame_000080_crop_352x192.png",
+ "sha256": "9496ad1c86063738c21ac8408ca1b09a50e650a3859a411ff0b6d7e8206b9d04",
+ "bytes": 15974
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/reference/frame_000000_crop_352x192.png",
+ "sha256": "99bb45d0363cf0f1adc9caffcfada05bd37cc4b325017cff0e130aef3c82a10d",
+ "bytes": 17007
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/reference/frame_000027_crop_352x192.png",
+ "sha256": "e04c26734a0ecb2a196f516c09c165db24159322bac27e1df230eb23b0079899",
+ "bytes": 17163
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/reference/frame_000053_crop_352x192.png",
+ "sha256": "2b6d8ca05917cc38bf5a9b329f58383aff57cfef658c40f50dc4447a31b07fdd",
+ "bytes": 16421
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/reference/frame_000080_crop_352x192.png",
+ "sha256": "1101fc930dfb45c44c66d190d1eda3ce7ef1b324b564fc9f00ea724075f399df",
+ "bytes": 16755
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/ours/frame_000000_crop_352x192.png",
+ "sha256": "bee0ef3bb44d136edc94431cebcfe0fb306d9a97090356299049461b971755aa",
+ "bytes": 16854
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/ours/frame_000027_crop_352x192.png",
+ "sha256": "55c9d4905d41b8575e7605db4918703a7af805c1b031c8cea2dcaf9ba1473f2a",
+ "bytes": 24655
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/ours/frame_000053_crop_352x192.png",
+ "sha256": "f4f872cb01653f08b4c82d4afff5508797c7a8eecf7ca5d513cafd19641c7984",
+ "bytes": 27805
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/ours/frame_000080_crop_352x192.png",
+ "sha256": "18395f9115854def64fb73efd411ac893f2b397beb9f1b61254a3d5ed15241ac",
+ "bytes": 30716
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/expected/frame_000000_crop_352x192.png",
+ "sha256": "97e87e204098390bd36bcb910286461fb0a8c02a1d0d67cf35e3b448eb1d988d",
+ "bytes": 16914
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/expected/frame_000027_crop_352x192.png",
+ "sha256": "3587ce8ed1dbfa12ff6f5d88e60912b63466a35fa98830efc3440b8db4710b56",
+ "bytes": 16975
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/expected/frame_000053_crop_352x192.png",
+ "sha256": "5c2e7ec7aec684019b7339544412f47025cc39b429d0a7acf418d6fa9bfb70b3",
+ "bytes": 16362
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/expected/frame_000080_crop_352x192.png",
+ "sha256": "6a6b18a984a86049e20c2736814d9de8080ae1322cfd966935e735fdfa5733d0",
+ "bytes": 16498
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/01_Source/frame_000000_crop_352x192.png",
+ "sha256": "97e87e204098390bd36bcb910286461fb0a8c02a1d0d67cf35e3b448eb1d988d",
+ "bytes": 16914
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/01_Source/frame_000027_crop_352x192.png",
+ "sha256": "f04beb2e5692267ac7ab00996702c4f944cc0b915e661f1a4fce2bc02ea44ad8",
+ "bytes": 17589
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/01_Source/frame_000053_crop_352x192.png",
+ "sha256": "d4d994c5b38330058cbae68a54178c0ee5e43866a956678cbea5a64671b486b6",
+ "bytes": 17737
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/01_Source/frame_000080_crop_352x192.png",
+ "sha256": "436c555af07fe825e9a67c208ba88ca36faa9ae62752977ac09a40c4ec5fa464",
+ "bytes": 18169
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/05_Learned_Reconstruction/frame_000000_crop_352x192.png",
+ "sha256": "e0f04e5641888dfc634e6a6b0756317f3394337c645203efad424dea24d73d77",
+ "bytes": 16850
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/05_Learned_Reconstruction/frame_000027_crop_352x192.png",
+ "sha256": "81f772fbe9cba6ddcbc0aa47a079d3abf935154b0eb4198a815b39ec55deb3fa",
+ "bytes": 23605
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/05_Learned_Reconstruction/frame_000053_crop_352x192.png",
+ "sha256": "9cb77dd9dcbf5f4ffd03db62059834c459f1e245c11a9f10e915bf7b990dff80",
+ "bytes": 26274
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/05_Learned_Reconstruction/frame_000080_crop_352x192.png",
+ "sha256": "3505eb81ef61e39c8d294aad3a512c3a6a53c0ec0741da5467e7ecf319ce2005",
+ "bytes": 29425
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/source.mp4",
+ "sha256": "c5802bfc0d735743654b44a284c16ce12963c20727f6569376d829d193d338b2",
+ "bytes": 1326039
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/reference.mp4",
+ "sha256": "986bf4e00eb44efaacaec63ec435dcfcfeb2fb24b4b211589367f0346c6f1206",
+ "bytes": 1331765
+ },
+ {
+ "asset": "../qualitative/synthetic/c399/ours.mp4",
+ "sha256": "bee49ee36fc1761488ecc772b5409881f93d8e3dd46a67f98ec4814124f4673c",
+ "bytes": 101970
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/source/frame_000000_crop_352x192.png",
+ "sha256": "038001835294506ae1c08d94fda86c4d8a1a62f71b5912dfdd518081005607cc",
+ "bytes": 17118
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/source/frame_000027_crop_352x192.png",
+ "sha256": "84563428581d487b48988dd428db9da7a413bdd91f6ea60ad5db4b0544f875b9",
+ "bytes": 18592
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/source/frame_000053_crop_352x192.png",
+ "sha256": "7478bd713230ac72e82dde939331eb9dcf12cf5ce113008ae3506e7a5518dea4",
+ "bytes": 18054
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/source/frame_000080_crop_352x192.png",
+ "sha256": "3e3f2ab4c0d01e36f0a04b93ccae814a88343554b6848faa93601eff120ce175",
+ "bytes": 17632
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/reference/frame_000000_crop_352x192.png",
+ "sha256": "99bb45d0363cf0f1adc9caffcfada05bd37cc4b325017cff0e130aef3c82a10d",
+ "bytes": 17007
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/reference/frame_000027_crop_352x192.png",
+ "sha256": "527d19ef73175770357369b3fefc740a4e9e2ba1fc2a1ed1445fa4ebaab309cd",
+ "bytes": 17619
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/reference/frame_000053_crop_352x192.png",
+ "sha256": "9855a83e5b9fd2b95e8bd5cf58b9978624d467c2d25fb8c65d2d289e6e64a3b1",
+ "bytes": 17677
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/reference/frame_000080_crop_352x192.png",
+ "sha256": "26eeb32f43ef1ba2bfff8e4048c3082ff559ee048805be8e87a741df92ee089e",
+ "bytes": 18252
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/ours/frame_000000_crop_352x192.png",
+ "sha256": "82e647b4bee02d068067917f071b524fa127f464cff04785a78ba2426c68e04d",
+ "bytes": 16884
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/ours/frame_000027_crop_352x192.png",
+ "sha256": "c29765b78c77b0ea1637ba8dfe5fa927ffd1b0b207daf0359036d12b08cd9f1d",
+ "bytes": 24378
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/ours/frame_000053_crop_352x192.png",
+ "sha256": "ed18ac981ac80c9b45cdda1ea4ef0919b71961bafdaafc400179ef56991d84f9",
+ "bytes": 26853
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/ours/frame_000080_crop_352x192.png",
+ "sha256": "52afa65a7852ea81e9d354925312b1b53f90f69177b5434b5f646fc5d0d1b4a5",
+ "bytes": 28881
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/expected/frame_000000_crop_352x192.png",
+ "sha256": "97e87e204098390bd36bcb910286461fb0a8c02a1d0d67cf35e3b448eb1d988d",
+ "bytes": 16914
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/expected/frame_000027_crop_352x192.png",
+ "sha256": "f04beb2e5692267ac7ab00996702c4f944cc0b915e661f1a4fce2bc02ea44ad8",
+ "bytes": 17589
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/expected/frame_000053_crop_352x192.png",
+ "sha256": "d4d994c5b38330058cbae68a54178c0ee5e43866a956678cbea5a64671b486b6",
+ "bytes": 17737
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/expected/frame_000080_crop_352x192.png",
+ "sha256": "436c555af07fe825e9a67c208ba88ca36faa9ae62752977ac09a40c4ec5fa464",
+ "bytes": 18169
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/01_Source/frame_000000_crop_352x192.png",
+ "sha256": "97e87e204098390bd36bcb910286461fb0a8c02a1d0d67cf35e3b448eb1d988d",
+ "bytes": 16914
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/01_Source/frame_000027_crop_352x192.png",
+ "sha256": "f04beb2e5692267ac7ab00996702c4f944cc0b915e661f1a4fce2bc02ea44ad8",
+ "bytes": 17589
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/01_Source/frame_000053_crop_352x192.png",
+ "sha256": "d4d994c5b38330058cbae68a54178c0ee5e43866a956678cbea5a64671b486b6",
+ "bytes": 17737
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/01_Source/frame_000080_crop_352x192.png",
+ "sha256": "436c555af07fe825e9a67c208ba88ca36faa9ae62752977ac09a40c4ec5fa464",
+ "bytes": 18169
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/05_Learned_Reconstruction/frame_000000_crop_352x192.png",
+ "sha256": "ac074b2333b716925e6e95d0edd63223b77ba732c69d94dd786b3045bf0a1330",
+ "bytes": 16469
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/05_Learned_Reconstruction/frame_000027_crop_352x192.png",
+ "sha256": "b1ebc5412b64845d452810b74a0d6817f098d0f48fd4cbe25bc6c3e9bf62e722",
+ "bytes": 22847
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/05_Learned_Reconstruction/frame_000053_crop_352x192.png",
+ "sha256": "c7024bbbc831f7cc7fd39d8e36ca97044ebfad8fa1b0100ff51e82d25fd0f2e4",
+ "bytes": 25120
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/05_Learned_Reconstruction/frame_000080_crop_352x192.png",
+ "sha256": "c88f148a932687e03652114fd06e2912f350bfe22720f3b28cf8e4e87c2b09ce",
+ "bytes": 28975
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/source.mp4",
+ "sha256": "9878afdd048b2a7dd1fc153fc7ac565c8f70bc8b08dc71b3b898b7d2a5f1d0f7",
+ "bytes": 1343802
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/reference.mp4",
+ "sha256": "ac4e3b8e6c6739d0c771d07a0bc853fac0bfabb5fe6837bce5d0ba00a78fb602",
+ "bytes": 1342331
+ },
+ {
+ "asset": "../qualitative/synthetic/c523/ours.mp4",
+ "sha256": "52fc7e9820225367660eb5ed0626c273c1ffa61affc0055dbbb37aaf964aa334",
+ "bytes": 98959
+ },
+ {
+ "asset": "../paper/source/2_theory.tex",
+ "sha256": "cea2d74371d5ebc780faf1715c3ed386bbc5263309f84754389b2057df4f35d7",
+ "bytes": 9601
+ },
+ {
+ "asset": "../paper/source/3_method.tex",
+ "sha256": "5536821070969800450da1769973067858e0e7ff6a7f993de9ea69be3227d55c",
+ "bytes": 7278
+ },
+ {
+ "asset": "../paper/source/4_experiments.tex",
+ "sha256": "2d25fec1beb1d52b71edfcc65c8ff839fbd47920f03db3f203b672c147352456",
+ "bytes": 6485
+ },
+ {
+ "asset": "../paper/source/appendix_d_implementation.tex",
+ "sha256": "ad32576448a5c159e9c6615a9d560589c999ba1318f4687b6d542aa751465778",
+ "bytes": 30773
+ },
+ {
+ "asset": "../paper/source/appendix_e_synthetic_protocol.tex",
+ "sha256": "4a28dec0bbbd86de384ff2471c30d3f3e2e188855eb650553520f7a698127dad",
+ "bytes": 23941
+ },
+ {
+ "asset": "../paper/source/appendix_b_unified_observations.tex",
+ "sha256": "95a7866b782ff54a2a6c0c1bfbdfe0f1c59eb0dfcb2fc6af4e72ace1a857337b",
+ "bytes": 13300
+ },
+ {
+ "asset": "../paper/source/table2_generation_editing_gap.csv",
+ "sha256": "ef6a377783bc08a517af075c02c8357ed50d3f8d0396701b75a3e6024569a1a9",
+ "bytes": 209
+ },
+ {
+ "asset": "../paper/source/synthetic_qualitative_v1.md",
+ "sha256": "087efc17ca957f5249910882d36f64483634e2d2c14ebd797b724205e4151a11",
+ "bytes": 5312
+ },
+ {
+ "asset": "../paper/source/real_qualitative_v1.md",
+ "sha256": "2b49117c5354603b70e373ea5ac84fa9c28f93a614403f8d4bc82728e15bb743",
+ "bytes": 4250
+ },
+ {
+ "asset": "../paper/source/real_qualitative_4case_v1.md",
+ "sha256": "c00f4663388fa0022b309fc4262bedda896df68146042bb9c1bc98b7b5cd03ae",
+ "bytes": 3547
+ },
+ {
+ "asset": "../paper/action_support_recovery.pdf",
+ "sha256": "ca2f41208e7f974211cf673c6ed56e4df3fbb5986a099a94adb3a8d014c7b7a4",
+ "bytes": 15447
+ }
+]
diff --git a/projects/video-action-editing/qualitative/omni/download_audit.json b/projects/video-action-editing/qualitative/omni/download_audit.json
new file mode 100644
index 0000000..f570ee0
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/download_audit.json
@@ -0,0 +1,272 @@
+[
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/datasets/omni_temporal200/Temporal/Motion Attribute/Motion_frequency_change/6_1280x720.mp4",
+ "local": "media/Te_Motion_frequency_change_06/source.mp4",
+ "sha256": "89033eb8d081cf564f28ff41111f25fcd6bd7f41604c0f4d3e54bb6f950f5ee8",
+ "bytes": 3961122
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/zebra_reference_weak_supervision/zebra_u680_bernini_pseudoref_v1/normalized/Te_Motion_frequency_change_06.mp4",
+ "local": "media/Te_Motion_frequency_change_06/zebra.mp4",
+ "sha256": "99e7098bf8ca8f327d9acf1c72b3c88b82ca7e79c3b7fc22c0fd503f53a88f23",
+ "bytes": 1732969
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/frozen_bernini/bernini_base_40step_temporal200_v1/normalized/Te_Motion_frequency_change_06.mp4",
+ "local": "media/Te_Motion_frequency_change_06/frozen_bernini.mp4",
+ "sha256": "197736d1ded8d96cfe65d3fd3a8982b9866a60a19d6b5799bfd337258db95700",
+ "bytes": 1596752
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/omnivideo2_official_1p3b/official_1p3b_40step_temporal200_v1/normalized/Te_Motion_frequency_change_06.mp4",
+ "local": "media/Te_Motion_frequency_change_06/omnivideo2_official_1p3b.mp4",
+ "sha256": "0ec0232e51cbdbced06ee86e9c81712facbe845ba0689521e2f45547925dd786",
+ "bytes": 3388422
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/lucy_edit_1p1_dev_5b/official_f12c3ab_50step_temporal200_v1/normalized/Te_Motion_frequency_change_06.mp4",
+ "local": "media/Te_Motion_frequency_change_06/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "8cdbcf5381196d017811d9f74464977bc946bb670dd9fb07d60b46b95278ee31",
+ "bytes": 1926421
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/vace_wan21_1p3b_fullframe/official_574e6a7_fullframe_50step_temporal200_v1/normalized/Te_Motion_frequency_change_06.mp4",
+ "local": "media/Te_Motion_frequency_change_06/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "d54f0a371d7519fd205d5fb20f38520953477a5d6bd61d807e58d735861f4c4f",
+ "bytes": 3857817
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/liveedit_wan21_1p3b/official_f0269e2_4step_temporal200_v1/normalized/Te_Motion_frequency_change_06.mp4",
+ "local": "media/Te_Motion_frequency_change_06/liveedit_wan21_1p3b.mp4",
+ "sha256": "920905582914e55b249fb889a697a64f987e527e4583fb70f912d6ce4d664164",
+ "bytes": 2745445
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/kiwi_edit_5b_instruct_only/official_277f43b_50step_temporal200_v1/normalized/Te_Motion_frequency_change_06.mp4",
+ "local": "media/Te_Motion_frequency_change_06/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "69143741503ec5c260fb5682fba38a9ac0a8d95a23689f582828454821c4c746",
+ "bytes": 1506545
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/reco_ori_wan21_1p3b/official_d6b471e_50step_temporal200_v1/normalized/Te_Motion_frequency_change_06.mp4",
+ "local": "media/Te_Motion_frequency_change_06/reco_ori_wan21_1p3b.mp4",
+ "sha256": "e59014ce6f558d6fb175cac92be6a90e8af459a6bc05eb255878ce42821a007b",
+ "bytes": 2486265
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/datasets/omni_temporal200/Temporal/Temporal Composition/Motion_deletion/1_1280x720.mp4",
+ "local": "media/Te_Motion_deletion_01/source.mp4",
+ "sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "bytes": 1037462
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/zebra_reference_weak_supervision/zebra_u680_bernini_pseudoref_v1/normalized/Te_Motion_deletion_01.mp4",
+ "local": "media/Te_Motion_deletion_01/zebra.mp4",
+ "sha256": "1d4de93dce10739e46069ec9f269a92de765c2f970dfa85d8de42a1875640ddd",
+ "bytes": 584758
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/frozen_bernini/bernini_base_40step_temporal200_v1/normalized/Te_Motion_deletion_01.mp4",
+ "local": "media/Te_Motion_deletion_01/frozen_bernini.mp4",
+ "sha256": "cd28f95da224595c3b323af7042fe788a08cb033637e3b958144287dd92a241e",
+ "bytes": 645612
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/omnivideo2_official_1p3b/official_1p3b_40step_temporal200_v1/normalized/Te_Motion_deletion_01.mp4",
+ "local": "media/Te_Motion_deletion_01/omnivideo2_official_1p3b.mp4",
+ "sha256": "362e2f0a8b72ccdac9fc6bb89eca33b5f54333d4b5454e7ba8c6ecf249aaa699",
+ "bytes": 1071023
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/lucy_edit_1p1_dev_5b/official_f12c3ab_50step_temporal200_v1/normalized/Te_Motion_deletion_01.mp4",
+ "local": "media/Te_Motion_deletion_01/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "bacf48ee36c1b5fad294c9ea3a3f6376792946c23297800b67f197ab3bd1d0be",
+ "bytes": 597859
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/vace_wan21_1p3b_fullframe/official_574e6a7_fullframe_50step_temporal200_v1/normalized/Te_Motion_deletion_01.mp4",
+ "local": "media/Te_Motion_deletion_01/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "c4a5537e65793a14c7f0aa816056105672f6864ff8b4bf89230f08a1ea28e371",
+ "bytes": 1414568
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/liveedit_wan21_1p3b/official_f0269e2_4step_temporal200_v1/normalized/Te_Motion_deletion_01.mp4",
+ "local": "media/Te_Motion_deletion_01/liveedit_wan21_1p3b.mp4",
+ "sha256": "b6758782a0ed97a2aae04121b61ad61fe69b481bae4a9220f4be2aeae381a889",
+ "bytes": 1194297
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/kiwi_edit_5b_instruct_only/official_277f43b_50step_temporal200_v1/normalized/Te_Motion_deletion_01.mp4",
+ "local": "media/Te_Motion_deletion_01/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "365a6e85a43a125865a6c94e892c32490234771de2cea0874c4ba9a7db803d50",
+ "bytes": 466006
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/reco_ori_wan21_1p3b/official_d6b471e_50step_temporal200_v1/normalized/Te_Motion_deletion_01.mp4",
+ "local": "media/Te_Motion_deletion_01/reco_ori_wan21_1p3b.mp4",
+ "sha256": "8febc79000879f423a5236cff8d712953999a152911f8d606c50bff0d458920f",
+ "bytes": 958841
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/datasets/omni_temporal200/Temporal/Motion Attribute/Motion_frequency_change/2_720x1280.mp4",
+ "local": "media/Te_Motion_frequency_change_02/source.mp4",
+ "sha256": "eee3a86a3441fb2617dcc8323c157d39fc83dc2bd2c0edf1ea860cd6d9c903d1",
+ "bytes": 938527
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/zebra_reference_weak_supervision/zebra_u680_bernini_pseudoref_v1/normalized/Te_Motion_frequency_change_02.mp4",
+ "local": "media/Te_Motion_frequency_change_02/zebra.mp4",
+ "sha256": "20c872bafe12b8de1073240051242732936beb02b3a2ea716fa18a5194ff70de",
+ "bytes": 831981
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/frozen_bernini/bernini_base_40step_temporal200_v1/normalized/Te_Motion_frequency_change_02.mp4",
+ "local": "media/Te_Motion_frequency_change_02/frozen_bernini.mp4",
+ "sha256": "e166a44a9312219d78537f13f49e8e04800a1814f02290ea39334ff7efddafd9",
+ "bytes": 881279
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/omnivideo2_official_1p3b/official_1p3b_40step_temporal200_v1/normalized/Te_Motion_frequency_change_02.mp4",
+ "local": "media/Te_Motion_frequency_change_02/omnivideo2_official_1p3b.mp4",
+ "sha256": "d030b29c60403f5c51b9c8d9272eb81853e270f81199ec05918b1d9db971dc78",
+ "bytes": 1182261
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/lucy_edit_1p1_dev_5b/official_f12c3ab_50step_temporal200_v1/normalized/Te_Motion_frequency_change_02.mp4",
+ "local": "media/Te_Motion_frequency_change_02/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "7faa97ec69c6559f1251d072e506a877e8e85f52239893bebd3a33d66b6269db",
+ "bytes": 749001
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/vace_wan21_1p3b_fullframe/official_574e6a7_fullframe_50step_temporal200_v1/normalized/Te_Motion_frequency_change_02.mp4",
+ "local": "media/Te_Motion_frequency_change_02/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "4fc6bd5fdcedb2703a879b09cc0460dc9da0646ab8d9774fa4ba828f46a4e972",
+ "bytes": 1345507
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/liveedit_wan21_1p3b/official_f0269e2_4step_temporal200_v1/normalized/Te_Motion_frequency_change_02.mp4",
+ "local": "media/Te_Motion_frequency_change_02/liveedit_wan21_1p3b.mp4",
+ "sha256": "a37e2101b84e691a17865a185c5900287036f1443d241d929a0fbe8c65180080",
+ "bytes": 967998
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/kiwi_edit_5b_instruct_only/official_277f43b_50step_temporal200_v1/normalized/Te_Motion_frequency_change_02.mp4",
+ "local": "media/Te_Motion_frequency_change_02/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "9fc6f9989ff621d63f3288406708eed133a514e7a5f5668eea57832e82ada495",
+ "bytes": 529899
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/reco_ori_wan21_1p3b/official_d6b471e_50step_temporal200_v1/normalized/Te_Motion_frequency_change_02.mp4",
+ "local": "media/Te_Motion_frequency_change_02/reco_ori_wan21_1p3b.mp4",
+ "sha256": "90db4129783fa802f016a01e0bd78b02327445d4a423c3fb499b103d2b75c21d",
+ "bytes": 854641
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/datasets/omni_temporal200/Temporal/Camera Attribute/Camera_speed_change/4_1280x720.mp4",
+ "local": "media/Te_Camera_speed_change_04/source.mp4",
+ "sha256": "0fddc8d23c7f6f5d55658de1ee5de4d588c20a78b6acf0344a7f467a5f6e6b84",
+ "bytes": 1393842
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/zebra_reference_weak_supervision/zebra_u680_bernini_pseudoref_v1/normalized/Te_Camera_speed_change_04.mp4",
+ "local": "media/Te_Camera_speed_change_04/zebra.mp4",
+ "sha256": "72a872773f4073862d9b67051d135f9f587bce0c2ca844fe4d918c8ed5e8c8de",
+ "bytes": 1147466
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/frozen_bernini/bernini_base_40step_temporal200_v1/normalized/Te_Camera_speed_change_04.mp4",
+ "local": "media/Te_Camera_speed_change_04/frozen_bernini.mp4",
+ "sha256": "4da6ebd8b81e975b4cdf8f7532c3c36b84dc792958e72c4bdb3c77653c7a3392",
+ "bytes": 1692502
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/omnivideo2_official_1p3b/official_1p3b_40step_temporal200_v1/normalized/Te_Camera_speed_change_04.mp4",
+ "local": "media/Te_Camera_speed_change_04/omnivideo2_official_1p3b.mp4",
+ "sha256": "7e909aa8b0a5bd039c10999b362a1f5de10926de3be2e8a4859e8b60d63a349b",
+ "bytes": 2969904
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/lucy_edit_1p1_dev_5b/official_f12c3ab_50step_temporal200_v1/normalized/Te_Camera_speed_change_04.mp4",
+ "local": "media/Te_Camera_speed_change_04/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "5e4e29619c5a16d24b9f9caf1459a9f9295d34669588d839f2b9ad6bc39627df",
+ "bytes": 692403
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/vace_wan21_1p3b_fullframe/official_574e6a7_fullframe_50step_temporal200_v1/normalized/Te_Camera_speed_change_04.mp4",
+ "local": "media/Te_Camera_speed_change_04/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "4f1a179eb4ae4737ffe1e7555b0ddd76413895f53090bc333900cf619a51452b",
+ "bytes": 1816057
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/liveedit_wan21_1p3b/official_f0269e2_4step_temporal200_v1/normalized/Te_Camera_speed_change_04.mp4",
+ "local": "media/Te_Camera_speed_change_04/liveedit_wan21_1p3b.mp4",
+ "sha256": "6c92a2c7edc0ba610fcf39579754239b2996cd457be2d8133e311c26d5e9b451",
+ "bytes": 1790100
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/kiwi_edit_5b_instruct_only/official_277f43b_50step_temporal200_v1/normalized/Te_Camera_speed_change_04.mp4",
+ "local": "media/Te_Camera_speed_change_04/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "a47894d39923fd8907bf97af4b40b1b24bcb4b2bccf4090e11ea99ea2972faa9",
+ "bytes": 1044453
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/reco_ori_wan21_1p3b/official_d6b471e_50step_temporal200_v1/normalized/Te_Camera_speed_change_04.mp4",
+ "local": "media/Te_Camera_speed_change_04/reco_ori_wan21_1p3b.mp4",
+ "sha256": "765f0c272a9f489fce69bdfa46a1d3826ad52af4832e031eb80c1a9f8696bb9e",
+ "bytes": 1865906
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/datasets/omni_temporal200/Temporal/Motion Semantic/Facial_expression_change/6_1280x720.mp4",
+ "local": "media/Te_Facial_expression_change_06/source.mp4",
+ "sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "bytes": 1655362
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/zebra_reference_weak_supervision/zebra_u680_bernini_pseudoref_v1/normalized/Te_Facial_expression_change_06.mp4",
+ "local": "media/Te_Facial_expression_change_06/zebra.mp4",
+ "sha256": "231b466565101154bf9185a812dbffa1195988d8873b746f82d393c913ad66a1",
+ "bytes": 1491176
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/frozen_bernini/bernini_base_40step_temporal200_v1/normalized/Te_Facial_expression_change_06.mp4",
+ "local": "media/Te_Facial_expression_change_06/frozen_bernini.mp4",
+ "sha256": "2234842811d1243ecb892b5beeb78ce3ef26fc0222a9f92cf067e59fede4fd2c",
+ "bytes": 1696848
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/omnivideo2_official_1p3b/official_1p3b_40step_temporal200_v1/normalized/Te_Facial_expression_change_06.mp4",
+ "local": "media/Te_Facial_expression_change_06/omnivideo2_official_1p3b.mp4",
+ "sha256": "95271b71d0cc214d83de53f2946b89778e77e076e26b4bc6a715372cfd5e1fa7",
+ "bytes": 3021981
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/lucy_edit_1p1_dev_5b/official_f12c3ab_50step_temporal200_v1/normalized/Te_Facial_expression_change_06.mp4",
+ "local": "media/Te_Facial_expression_change_06/lucy_edit_1p1_dev_5b.mp4",
+ "sha256": "f0285123defe401fe6d9b707716de369dc8162402f64a59748a756752dfb0e2b",
+ "bytes": 825523
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/vace_wan21_1p3b_fullframe/official_574e6a7_fullframe_50step_temporal200_v1/normalized/Te_Facial_expression_change_06.mp4",
+ "local": "media/Te_Facial_expression_change_06/vace_wan21_1p3b_fullframe.mp4",
+ "sha256": "f2194f1c764e6f1f00b1e9e59c0c213e4e574dfa3024cef0d565ba165d081748",
+ "bytes": 1597654
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/liveedit_wan21_1p3b/official_f0269e2_4step_temporal200_v1/normalized/Te_Facial_expression_change_06.mp4",
+ "local": "media/Te_Facial_expression_change_06/liveedit_wan21_1p3b.mp4",
+ "sha256": "8ba93c48945cd80e3248682620976249e165313f5042c4a7ec04166ee8a28ddf",
+ "bytes": 2023071
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/kiwi_edit_5b_instruct_only/official_277f43b_50step_temporal200_v1/normalized/Te_Facial_expression_change_06.mp4",
+ "local": "media/Te_Facial_expression_change_06/kiwi_edit_5b_instruct_only.mp4",
+ "sha256": "3dbb96889496be16482f84a759a34a4e38e39dbde4a376b4424054efd06ea740",
+ "bytes": 1085868
+ },
+ {
+ "path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VideoEditing/VideoEdit_experiments/action_editing_benchmark_v1_20260911/runs/omni_temporal200/reco_ori_wan21_1p3b/official_d6b471e_50step_temporal200_v1/normalized/Te_Facial_expression_change_06.mp4",
+ "local": "media/Te_Facial_expression_change_06/reco_ori_wan21_1p3b.mp4",
+ "sha256": "a23c0cb3676dc3904957c80dad444a59651862972ecc31155bc01fd40d525246",
+ "bytes": 865431
+ }
+]
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/frozen_bernini.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/frozen_bernini.json
new file mode 100644
index 0000000..2ea9a75
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/frozen_bernini.json
@@ -0,0 +1,46 @@
+{
+ "category": "Camera Attribute",
+ "created_at_utc": "2026-09-14T18:10:10Z",
+ "edited_video_sha256": "4da6ebd8b81e975b4cdf8f7532c3c36b84dc792958e72c4bdb3c77653c7a3392",
+ "editing_instruction": "Accelerate the camera's movement speed.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "35bd76d4ea58ddf315ea3630265a781fca041769aa67c81f839b44de7a7e9141",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 63,
+ "item_id": "Te_Camera_speed_change_04",
+ "parsed_judge_output": {
+ "accuracy": 5,
+ "consistency": 4,
+ "preservation": 3,
+ "realism": 4,
+ "reason": "The camera movement speed is significantly accelerated, as instructed, with smooth and continuous motion toward the skyscraper. The action is correctly executed. However, preservation is reduced due to noticeable changes in lighting, color grading, and sky details (e.g., clouds appear more dynamic and stylized) that were not part of the original video. Realism is high with natural motion, and consistency is good with no temporal drift or artifacts. The edit is visually enhanced but not strictly preserving original content."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 5,\n \"preservation\": 3,\n \"realism\": 4,\n \"consistency\": 4,\n \"reason\": \"The camera movement speed is significantly accelerated, as instructed, with smooth and continuous motion toward the skyscraper. The action is correctly executed. However, preservation is reduced due to noticeable changes in lighting, color grading, and sky details (e.g., clouds appear more dynamic and stylized) that were not part of the original video. Realism is high with natural motion, and consistency is good with no temporal drift or artifacts. The edit is visually enhanced but not strictly preserving original content.\"\n}",
+ "receipt_digest": "4d8d16e68a0d9936562900a9a5b749b96c55abf2ee5d2a44eb6d575ff036b0ce",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 5.0,
+ "accuracy_penalty": 1.0,
+ "consistency": 4.0,
+ "overall_score_paper_0_to_100": 86.0,
+ "overall_score_raw_0_to_5": 4.3,
+ "preservation": 3.0,
+ "realism": 4.0,
+ "reason": "The camera movement speed is significantly accelerated, as instructed, with smooth and continuous motion toward the skyscraper. The action is correctly executed. However, preservation is reduced due to noticeable changes in lighting, color grading, and sky details (e.g., clouds appear more dynamic and stylized) that were not part of the original video. Realism is high with natural motion, and consistency is good with no temporal drift or artifacts. The edit is visually enhanced but not strictly preserving original content."
+ },
+ "source_video_sha256": "0fddc8d23c7f6f5d55658de1ee5de4d588c20a78b6acf0344a7f467a5f6e6b84",
+ "subcategory": "Camera_speed_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/kiwi_edit_5b_instruct_only.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/kiwi_edit_5b_instruct_only.json
new file mode 100644
index 0000000..dc5c087
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/kiwi_edit_5b_instruct_only.json
@@ -0,0 +1,46 @@
+{
+ "category": "Camera Attribute",
+ "created_at_utc": "2026-09-14T20:02:16Z",
+ "edited_video_sha256": "a47894d39923fd8907bf97af4b40b1b24bcb4b2bccf4090e11ea99ea2972faa9",
+ "editing_instruction": "Accelerate the camera's movement speed.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "0329899b19a0300bc281bd54a128d9314004f9ffe4eadff295b2fee6ed5195aa",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 63,
+ "item_id": "Te_Camera_speed_change_04",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 2,
+ "preservation": 3,
+ "realism": 2,
+ "reason": "The instruction was to accelerate the camera's movement speed. However, the edited video shows no noticeable change in camera speed compared to the original. The camera movement appears identical in both videos, indicating the edit failed to achieve the requested acceleration. Additionally, there are visual artifacts and distortions on the central skyscraper (e.g., pixelation and color bleeding), suggesting a failed or incorrect edit. Preservation is partially maintained as background elements remain mostly unchanged, but realism and consistency are compromised due to these artifacts and lack of temporal coherence."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 3,\n \"realism\": 2,\n \"consistency\": 2,\n \"reason\": \"The instruction was to accelerate the camera's movement speed. However, the edited video shows no noticeable change in camera speed compared to the original. The camera movement appears identical in both videos, indicating the edit failed to achieve the requested acceleration. Additionally, there are visual artifacts and distortions on the central skyscraper (e.g., pixelation and color bleeding), suggesting a failed or incorrect edit. Preservation is partially maintained as background elements remain mostly unchanged, but realism and consistency are compromised due to these artifacts and lack of temporal coherence.\"\n}",
+ "receipt_digest": "3ebcb95ef4ba1226f2037b803131d187d58b34b99543ca5f44c34694bc23fe67",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 2.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 3.0,
+ "realism": 2.0,
+ "reason": "The instruction was to accelerate the camera's movement speed. However, the edited video shows no noticeable change in camera speed compared to the original. The camera movement appears identical in both videos, indicating the edit failed to achieve the requested acceleration. Additionally, there are visual artifacts and distortions on the central skyscraper (e.g., pixelation and color bleeding), suggesting a failed or incorrect edit. Preservation is partially maintained as background elements remain mostly unchanged, but realism and consistency are compromised due to these artifacts and lack of temporal coherence."
+ },
+ "source_video_sha256": "0fddc8d23c7f6f5d55658de1ee5de4d588c20a78b6acf0344a7f467a5f6e6b84",
+ "subcategory": "Camera_speed_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/liveedit_wan21_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/liveedit_wan21_1p3b.json
new file mode 100644
index 0000000..b44c537
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/liveedit_wan21_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Camera Attribute",
+ "created_at_utc": "2026-09-14T19:58:25Z",
+ "edited_video_sha256": "6c92a2c7edc0ba610fcf39579754239b2996cd457be2d8133e311c26d5e9b451",
+ "editing_instruction": "Accelerate the camera's movement speed.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "fc731a49e6c3d35b5945782f04ebc463488a0e3e5ede54e972425835c208e5d7",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 63,
+ "item_id": "Te_Camera_speed_change_04",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 1,
+ "preservation": 2,
+ "realism": 1,
+ "reason": "The instruction was to accelerate the camera's movement speed, but the edited video shows no change in camera motion speed compared to the original. Instead, there is a black, distorted object superimposed on the tower, which is a major error. This object is not present in the original and appears to be an artifact or incorrect edit. The camera movement remains identical, so the core instruction is completely unmet. Preservation is low because unrelated content (the tower and city) is altered by the black object. Realism and consistency are poor due to the unnatural artifact and lack of temporal stability."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 1,\n \"consistency\": 1,\n \"reason\": \"The instruction was to accelerate the camera's movement speed, but the edited video shows no change in camera motion speed compared to the original. Instead, there is a black, distorted object superimposed on the tower, which is a major error. This object is not present in the original and appears to be an artifact or incorrect edit. The camera movement remains identical, so the core instruction is completely unmet. Preservation is low because unrelated content (the tower and city) is altered by the black object. Realism and consistency are poor due to the unnatural artifact and lack of temporal stability.\"\n}",
+ "receipt_digest": "bdd9006c9e3f951d28e279bcdbdfc664c02cda1bf7454b2a4b2cfe10689fd537",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 1.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 1.0,
+ "reason": "The instruction was to accelerate the camera's movement speed, but the edited video shows no change in camera motion speed compared to the original. Instead, there is a black, distorted object superimposed on the tower, which is a major error. This object is not present in the original and appears to be an artifact or incorrect edit. The camera movement remains identical, so the core instruction is completely unmet. Preservation is low because unrelated content (the tower and city) is altered by the black object. Realism and consistency are poor due to the unnatural artifact and lack of temporal stability."
+ },
+ "source_video_sha256": "0fddc8d23c7f6f5d55658de1ee5de4d588c20a78b6acf0344a7f467a5f6e6b84",
+ "subcategory": "Camera_speed_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/lucy_edit_1p1_dev_5b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/lucy_edit_1p1_dev_5b.json
new file mode 100644
index 0000000..8a04eba
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/lucy_edit_1p1_dev_5b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Camera Attribute",
+ "created_at_utc": "2026-09-14T18:07:26Z",
+ "edited_video_sha256": "5e4e29619c5a16d24b9f9caf1459a9f9295d34669588d839f2b9ad6bc39627df",
+ "editing_instruction": "Accelerate the camera's movement speed.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "958406a72abc3d1d3ccff777bbe657b4e77061cb2ecce850fc4a0f8546716c3a",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 63,
+ "item_id": "Te_Camera_speed_change_04",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 5,
+ "preservation": 5,
+ "realism": 5,
+ "reason": "The instruction was to accelerate the camera's movement speed. However, the edited video shows no noticeable change in camera speed compared to the original. The camera movement remains identical in timing and trajectory, indicating the edit failed to apply the requested acceleration. Since accuracy is the top priority and the instruction was not followed, accuracy is scored 0. Preservation, realism, and consistency are high as the rest of the video remains unchanged and visually coherent."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 5,\n \"realism\": 5,\n \"consistency\": 5,\n \"reason\": \"The instruction was to accelerate the camera's movement speed. However, the edited video shows no noticeable change in camera speed compared to the original. The camera movement remains identical in timing and trajectory, indicating the edit failed to apply the requested acceleration. Since accuracy is the top priority and the instruction was not followed, accuracy is scored 0. Preservation, realism, and consistency are high as the rest of the video remains unchanged and visually coherent.\"\n}",
+ "receipt_digest": "697c0c79b123d093326500f5b52cca85f5bdeae20eb9e9a8c6fd28b4d10ba2cf",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 5.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 5.0,
+ "realism": 5.0,
+ "reason": "The instruction was to accelerate the camera's movement speed. However, the edited video shows no noticeable change in camera speed compared to the original. The camera movement remains identical in timing and trajectory, indicating the edit failed to apply the requested acceleration. Since accuracy is the top priority and the instruction was not followed, accuracy is scored 0. Preservation, realism, and consistency are high as the rest of the video remains unchanged and visually coherent."
+ },
+ "source_video_sha256": "0fddc8d23c7f6f5d55658de1ee5de4d588c20a78b6acf0344a7f467a5f6e6b84",
+ "subcategory": "Camera_speed_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/omnivideo2_official_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/omnivideo2_official_1p3b.json
new file mode 100644
index 0000000..880ce69
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/omnivideo2_official_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Camera Attribute",
+ "created_at_utc": "2026-09-14T18:11:15Z",
+ "edited_video_sha256": "7e909aa8b0a5bd039c10999b362a1f5de10926de3be2e8a4859e8b60d63a349b",
+ "editing_instruction": "Accelerate the camera's movement speed.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "1b79bbf410c6adf53abd87332bc0cc6930b4f54494ede953234b3355e38afefb",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 63,
+ "item_id": "Te_Camera_speed_change_04",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 2,
+ "preservation": 0,
+ "realism": 2,
+ "reason": "The instruction was to accelerate the camera's movement speed, but the edited video shows no change in camera motion — the camera remains static, identical to the original. Additionally, the central skyscraper's facade has been altered (pixelated pattern replaced with smooth glass), which is a major structural change unrelated to the instruction, violating preservation. The edits are not realistic or consistent, as the building's appearance is inconsistent with the original and shows no temporal motion. Therefore, accuracy is 0 due to failure to follow the instruction, and preservation is 0 due to unintended content modification."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 0,\n \"realism\": 2,\n \"consistency\": 2,\n \"reason\": \"The instruction was to accelerate the camera's movement speed, but the edited video shows no change in camera motion — the camera remains static, identical to the original. Additionally, the central skyscraper's facade has been altered (pixelated pattern replaced with smooth glass), which is a major structural change unrelated to the instruction, violating preservation. The edits are not realistic or consistent, as the building's appearance is inconsistent with the original and shows no temporal motion. Therefore, accuracy is 0 due to failure to follow the instruction, and preservation is 0 due to unintended content modification.\"\n}",
+ "receipt_digest": "828f79255c938b339d68238c497f0a1e751f3ebda8cddfe7ee9badf8af19e833",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 2.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 0.0,
+ "realism": 2.0,
+ "reason": "The instruction was to accelerate the camera's movement speed, but the edited video shows no change in camera motion — the camera remains static, identical to the original. Additionally, the central skyscraper's facade has been altered (pixelated pattern replaced with smooth glass), which is a major structural change unrelated to the instruction, violating preservation. The edits are not realistic or consistent, as the building's appearance is inconsistent with the original and shows no temporal motion. Therefore, accuracy is 0 due to failure to follow the instruction, and preservation is 0 due to unintended content modification."
+ },
+ "source_video_sha256": "0fddc8d23c7f6f5d55658de1ee5de4d588c20a78b6acf0344a7f467a5f6e6b84",
+ "subcategory": "Camera_speed_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/reco_ori_wan21_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/reco_ori_wan21_1p3b.json
new file mode 100644
index 0000000..aa78bbc
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/reco_ori_wan21_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Camera Attribute",
+ "created_at_utc": "2026-09-14T19:56:21Z",
+ "edited_video_sha256": "765f0c272a9f489fce69bdfa46a1d3826ad52af4832e031eb80c1a9f8696bb9e",
+ "editing_instruction": "Accelerate the camera's movement speed.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "ce02857d7678e154c3e4ed0e6feab8073b9587fe0c17550586d549a5e1bf13e5",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 63,
+ "item_id": "Te_Camera_speed_change_04",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 2,
+ "preservation": 2,
+ "realism": 2,
+ "reason": "The instruction was to accelerate the camera's movement speed. However, the edited video shows no noticeable change in camera speed compared to the original. The camera movement appears identical in both videos, indicating the edit failed to achieve the intended effect. Additionally, the central skyscraper's top structure is visibly distorted or missing in the edited video, suggesting a major artifact or error in processing. This structural change is unrelated to the instruction and violates preservation. The motion is also slightly jittery and unnatural, reducing realism and consistency."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 2,\n \"consistency\": 2,\n \"reason\": \"The instruction was to accelerate the camera's movement speed. However, the edited video shows no noticeable change in camera speed compared to the original. The camera movement appears identical in both videos, indicating the edit failed to achieve the intended effect. Additionally, the central skyscraper's top structure is visibly distorted or missing in the edited video, suggesting a major artifact or error in processing. This structural change is unrelated to the instruction and violates preservation. The motion is also slightly jittery and unnatural, reducing realism and consistency.\"\n}",
+ "receipt_digest": "baa3a9b3654ed647062f8e26127e53ef0a29cba1a94b74914ca54745ce5af0f5",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 2.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 2.0,
+ "reason": "The instruction was to accelerate the camera's movement speed. However, the edited video shows no noticeable change in camera speed compared to the original. The camera movement appears identical in both videos, indicating the edit failed to achieve the intended effect. Additionally, the central skyscraper's top structure is visibly distorted or missing in the edited video, suggesting a major artifact or error in processing. This structural change is unrelated to the instruction and violates preservation. The motion is also slightly jittery and unnatural, reducing realism and consistency."
+ },
+ "source_video_sha256": "0fddc8d23c7f6f5d55658de1ee5de4d588c20a78b6acf0344a7f467a5f6e6b84",
+ "subcategory": "Camera_speed_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/vace_wan21_1p3b_fullframe.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/vace_wan21_1p3b_fullframe.json
new file mode 100644
index 0000000..117ea47
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/vace_wan21_1p3b_fullframe.json
@@ -0,0 +1,46 @@
+{
+ "category": "Camera Attribute",
+ "created_at_utc": "2026-09-14T19:46:36Z",
+ "edited_video_sha256": "4f1a179eb4ae4737ffe1e7555b0ddd76413895f53090bc333900cf619a51452b",
+ "editing_instruction": "Accelerate the camera's movement speed.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "dd144add239638ebbc6a7f5dd0112fdc48e4b0eb193048b94cd97eecdeacb179",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 63,
+ "item_id": "Te_Camera_speed_change_04",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 3,
+ "preservation": 2,
+ "realism": 3,
+ "reason": "The instruction was to accelerate the camera's movement speed, but the edited video shows no noticeable change in camera motion speed compared to the original. The camera movement appears identical in timing and trajectory, indicating the edit failed to fulfill the core instruction. Additionally, there is a significant color grading change (warmer, more saturated tones) which is an unintended modification to unrelated content, violating preservation. While motion is smooth and consistent, the failure to accelerate the camera movement makes accuracy 0."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 3,\n \"consistency\": 3,\n \"reason\": \"The instruction was to accelerate the camera's movement speed, but the edited video shows no noticeable change in camera motion speed compared to the original. The camera movement appears identical in timing and trajectory, indicating the edit failed to fulfill the core instruction. Additionally, there is a significant color grading change (warmer, more saturated tones) which is an unintended modification to unrelated content, violating preservation. While motion is smooth and consistent, the failure to accelerate the camera movement makes accuracy 0.\"\n}",
+ "receipt_digest": "89b89a4995815d6747f0c84ba24fc2a5b006edcc2fd10da3cf925935f41bf600",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 3.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 3.0,
+ "reason": "The instruction was to accelerate the camera's movement speed, but the edited video shows no noticeable change in camera motion speed compared to the original. The camera movement appears identical in timing and trajectory, indicating the edit failed to fulfill the core instruction. Additionally, there is a significant color grading change (warmer, more saturated tones) which is an unintended modification to unrelated content, violating preservation. While motion is smooth and consistent, the failure to accelerate the camera movement makes accuracy 0."
+ },
+ "source_video_sha256": "0fddc8d23c7f6f5d55658de1ee5de4d588c20a78b6acf0344a7f467a5f6e6b84",
+ "subcategory": "Camera_speed_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/zebra.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/zebra.json
new file mode 100644
index 0000000..0e2df9b
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Camera_speed_change_04/zebra.json
@@ -0,0 +1,46 @@
+{
+ "category": "Camera Attribute",
+ "created_at_utc": "2026-09-22T09:56:19Z",
+ "edited_video_sha256": "72a872773f4073862d9b67051d135f9f587bce0c2ca844fe4d918c8ed5e8c8de",
+ "editing_instruction": "Accelerate the camera's movement speed.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "149da52dc20cd8ac0dcb568d1eb13e785bc243ac1c0d1be8003e09e0e9938f7b",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 63,
+ "item_id": "Te_Camera_speed_change_04",
+ "parsed_judge_output": {
+ "accuracy": 5,
+ "consistency": 5,
+ "preservation": 5,
+ "realism": 5,
+ "reason": "The camera movement speed is clearly accelerated in the edited video compared to the original, as evidenced by the faster panning motion and quicker reveal of the cityscape. The movement is smooth and natural, with no visible artifacts. All other elements (buildings, sky, water, mountains) remain unchanged and consistent across frames, preserving the original content. Temporal and spatial consistency is maintained throughout the video. The edit correctly follows the instruction with high fidelity."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 5,\n \"preservation\": 5,\n \"realism\": 5,\n \"consistency\": 5,\n \"reason\": \"The camera movement speed is clearly accelerated in the edited video compared to the original, as evidenced by the faster panning motion and quicker reveal of the cityscape. The movement is smooth and natural, with no visible artifacts. All other elements (buildings, sky, water, mountains) remain unchanged and consistent across frames, preserving the original content. Temporal and spatial consistency is maintained throughout the video. The edit correctly follows the instruction with high fidelity.\"\n}",
+ "receipt_digest": "94ed80ca792091a9090a299dd2732299a71efb4a12b8a43d15aa8bdbf8c43253",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 5.0,
+ "accuracy_penalty": 1.0,
+ "consistency": 5.0,
+ "overall_score_paper_0_to_100": 100.0,
+ "overall_score_raw_0_to_5": 5.0,
+ "preservation": 5.0,
+ "realism": 5.0,
+ "reason": "The camera movement speed is clearly accelerated in the edited video compared to the original, as evidenced by the faster panning motion and quicker reveal of the cityscape. The movement is smooth and natural, with no visible artifacts. All other elements (buildings, sky, water, mountains) remain unchanged and consistent across frames, preserving the original content. Temporal and spatial consistency is maintained throughout the video. The edit correctly follows the instruction with high fidelity."
+ },
+ "source_video_sha256": "0fddc8d23c7f6f5d55658de1ee5de4d588c20a78b6acf0344a7f467a5f6e6b84",
+ "subcategory": "Camera_speed_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/frozen_bernini.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/frozen_bernini.json
new file mode 100644
index 0000000..ffa0ce3
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/frozen_bernini.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Semantic",
+ "created_at_utc": "2026-09-14T17:50:52Z",
+ "edited_video_sha256": "2234842811d1243ecb892b5beeb78ce3ef26fc0222a9f92cf067e59fede4fd2c",
+ "editing_instruction": "Make the man laugh happily.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "35bd76d4ea58ddf315ea3630265a781fca041769aa67c81f839b44de7a7e9141",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 25,
+ "item_id": "Te_Facial_expression_change_06",
+ "parsed_judge_output": {
+ "accuracy": 5,
+ "consistency": 4,
+ "preservation": 3,
+ "realism": 4,
+ "reason": "The instruction to make the man laugh happily is fully executed: his facial expression transitions naturally into a wide, joyful laugh with crinkled eyes and visible teeth, matching the emotional intent. However, background elements (e.g., lighting, foliage, and structures) show noticeable changes between original and edited video, indicating some preservation issues. The motion of the laugh is smooth and realistic, with no visible artifacts. Temporal consistency is mostly maintained, though slight shifts in background depth of field and lighting suggest minor inconsistencies. Overall, the edit is accurate and realistic, but background preservation is imperfect."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 5,\n \"preservation\": 3,\n \"realism\": 4,\n \"consistency\": 4,\n \"reason\": \"The instruction to make the man laugh happily is fully executed: his facial expression transitions naturally into a wide, joyful laugh with crinkled eyes and visible teeth, matching the emotional intent. However, background elements (e.g., lighting, foliage, and structures) show noticeable changes between original and edited video, indicating some preservation issues. The motion of the laugh is smooth and realistic, with no visible artifacts. Temporal consistency is mostly maintained, though slight shifts in background depth of field and lighting suggest minor inconsistencies. Overall, the edit is accurate and realistic, but background preservation is imperfect.\"\n}",
+ "receipt_digest": "3820058a146dceca4241b5e3f2c24a347a0195a2491e398b2b8661eae4773a2b",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 5.0,
+ "accuracy_penalty": 1.0,
+ "consistency": 4.0,
+ "overall_score_paper_0_to_100": 86.0,
+ "overall_score_raw_0_to_5": 4.3,
+ "preservation": 3.0,
+ "realism": 4.0,
+ "reason": "The instruction to make the man laugh happily is fully executed: his facial expression transitions naturally into a wide, joyful laugh with crinkled eyes and visible teeth, matching the emotional intent. However, background elements (e.g., lighting, foliage, and structures) show noticeable changes between original and edited video, indicating some preservation issues. The motion of the laugh is smooth and realistic, with no visible artifacts. Temporal consistency is mostly maintained, though slight shifts in background depth of field and lighting suggest minor inconsistencies. Overall, the edit is accurate and realistic, but background preservation is imperfect."
+ },
+ "source_video_sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "subcategory": "Facial_expression_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/kiwi_edit_5b_instruct_only.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/kiwi_edit_5b_instruct_only.json
new file mode 100644
index 0000000..53f9f96
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/kiwi_edit_5b_instruct_only.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Semantic",
+ "created_at_utc": "2026-09-14T19:42:03Z",
+ "edited_video_sha256": "3dbb96889496be16482f84a759a34a4e38e39dbde4a376b4424054efd06ea740",
+ "editing_instruction": "Make the man laugh happily.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "0329899b19a0300bc281bd54a128d9314004f9ffe4eadff295b2fee6ed5195aa",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 25,
+ "item_id": "Te_Facial_expression_change_06",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 4,
+ "preservation": 3,
+ "realism": 4,
+ "reason": "The instruction was to make the man laugh happily, but the edited video shows no change in expression—his face remains neutral or slightly pensive, with no signs of laughter (e.g., raised cheeks, crinkled eyes, open mouth). The edit fails to alter the intended emotion. Background and lighting are preserved reasonably well, and motion is smooth, but the core instruction is not followed, so accuracy is 0."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 3,\n \"realism\": 4,\n \"consistency\": 4,\n \"reason\": \"The instruction was to make the man laugh happily, but the edited video shows no change in expression—his face remains neutral or slightly pensive, with no signs of laughter (e.g., raised cheeks, crinkled eyes, open mouth). The edit fails to alter the intended emotion. Background and lighting are preserved reasonably well, and motion is smooth, but the core instruction is not followed, so accuracy is 0.\"\n}",
+ "receipt_digest": "f7804f976165fc8971733504287e95fc0b048863047a17bc3fc80e000b7b29f7",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 4.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 3.0,
+ "realism": 4.0,
+ "reason": "The instruction was to make the man laugh happily, but the edited video shows no change in expression—his face remains neutral or slightly pensive, with no signs of laughter (e.g., raised cheeks, crinkled eyes, open mouth). The edit fails to alter the intended emotion. Background and lighting are preserved reasonably well, and motion is smooth, but the core instruction is not followed, so accuracy is 0."
+ },
+ "source_video_sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "subcategory": "Facial_expression_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/liveedit_wan21_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/liveedit_wan21_1p3b.json
new file mode 100644
index 0000000..fdd4830
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/liveedit_wan21_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Semantic",
+ "created_at_utc": "2026-09-14T19:37:29Z",
+ "edited_video_sha256": "8ba93c48945cd80e3248682620976249e165313f5042c4a7ec04166ee8a28ddf",
+ "editing_instruction": "Make the man laugh happily.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "fc731a49e6c3d35b5945782f04ebc463488a0e3e5ede54e972425835c208e5d7",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 25,
+ "item_id": "Te_Facial_expression_change_06",
+ "parsed_judge_output": {
+ "accuracy": 2,
+ "consistency": 3,
+ "preservation": 3,
+ "realism": 2,
+ "reason": "The instruction was to make the man laugh happily. While the man is shown with a wide smile and crinkled eyes, the expression appears unnatural and exaggerated, with overly bright, uniform teeth and unnaturally pink lips, suggesting digital manipulation rather than a genuine laugh. The eyes also appear to shift unnaturally, and the skin texture is overly smooth and plastic-like, indicating poor realism. The background is preserved reasonably well, but there are slight inconsistencies in lighting and depth of field. Temporal consistency is acceptable, but the facial animation lacks natural motion. Overall, the edit partially achieves the goal but fails in realism and natural expression."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 2,\n \"preservation\": 3,\n \"realism\": 2,\n \"consistency\": 3,\n \"reason\": \"The instruction was to make the man laugh happily. While the man is shown with a wide smile and crinkled eyes, the expression appears unnatural and exaggerated, with overly bright, uniform teeth and unnaturally pink lips, suggesting digital manipulation rather than a genuine laugh. The eyes also appear to shift unnaturally, and the skin texture is overly smooth and plastic-like, indicating poor realism. The background is preserved reasonably well, but there are slight inconsistencies in lighting and depth of field. Temporal consistency is acceptable, but the facial animation lacks natural motion. Overall, the edit partially achieves the goal but fails in realism and natural expression.\"\n}",
+ "receipt_digest": "e964bb54b7781f59507bdad87538ef3b3cf76e9fa975c983f60cb9e99271cf0d",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 2.0,
+ "accuracy_penalty": 0.4,
+ "consistency": 3.0,
+ "overall_score_paper_0_to_100": 30.8,
+ "overall_score_raw_0_to_5": 1.54,
+ "preservation": 3.0,
+ "realism": 2.0,
+ "reason": "The instruction was to make the man laugh happily. While the man is shown with a wide smile and crinkled eyes, the expression appears unnatural and exaggerated, with overly bright, uniform teeth and unnaturally pink lips, suggesting digital manipulation rather than a genuine laugh. The eyes also appear to shift unnaturally, and the skin texture is overly smooth and plastic-like, indicating poor realism. The background is preserved reasonably well, but there are slight inconsistencies in lighting and depth of field. Temporal consistency is acceptable, but the facial animation lacks natural motion. Overall, the edit partially achieves the goal but fails in realism and natural expression."
+ },
+ "source_video_sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "subcategory": "Facial_expression_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/lucy_edit_1p1_dev_5b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/lucy_edit_1p1_dev_5b.json
new file mode 100644
index 0000000..a71a154
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/lucy_edit_1p1_dev_5b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Semantic",
+ "created_at_utc": "2026-09-14T17:48:15Z",
+ "edited_video_sha256": "f0285123defe401fe6d9b707716de369dc8162402f64a59748a756752dfb0e2b",
+ "editing_instruction": "Make the man laugh happily.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "958406a72abc3d1d3ccff777bbe657b4e77061cb2ecce850fc4a0f8546716c3a",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 25,
+ "item_id": "Te_Facial_expression_change_06",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 5,
+ "preservation": 5,
+ "realism": 5,
+ "reason": "The instruction was to make the man laugh happily, but the edited video shows no change in his expression—he remains serious and neutral throughout. The edit failed to alter his facial expression to reflect laughter, making accuracy 0. Preservation, realism, and consistency are high as the background, lighting, and motion remain unchanged and natural."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 5,\n \"realism\": 5,\n \"consistency\": 5,\n \"reason\": \"The instruction was to make the man laugh happily, but the edited video shows no change in his expression—he remains serious and neutral throughout. The edit failed to alter his facial expression to reflect laughter, making accuracy 0. Preservation, realism, and consistency are high as the background, lighting, and motion remain unchanged and natural.\"\n}",
+ "receipt_digest": "826301c84771323042b8edccebadad79594db86621eea85ec1a225012fb9f392",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 5.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 5.0,
+ "realism": 5.0,
+ "reason": "The instruction was to make the man laugh happily, but the edited video shows no change in his expression—he remains serious and neutral throughout. The edit failed to alter his facial expression to reflect laughter, making accuracy 0. Preservation, realism, and consistency are high as the background, lighting, and motion remain unchanged and natural."
+ },
+ "source_video_sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "subcategory": "Facial_expression_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/omnivideo2_official_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/omnivideo2_official_1p3b.json
new file mode 100644
index 0000000..620a247
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/omnivideo2_official_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Semantic",
+ "created_at_utc": "2026-09-14T17:50:32Z",
+ "edited_video_sha256": "95271b71d0cc214d83de53f2946b89778e77e076e26b4bc6a715372cfd5e1fa7",
+ "editing_instruction": "Make the man laugh happily.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "1b79bbf410c6adf53abd87332bc0cc6930b4f54494ede953234b3355e38afefb",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 25,
+ "item_id": "Te_Facial_expression_change_06",
+ "parsed_judge_output": {
+ "accuracy": 5,
+ "consistency": 4,
+ "preservation": 3,
+ "realism": 4,
+ "reason": "The instruction to make the man laugh happily is fully executed: his facial expression shows a genuine, wide smile with crinkled eyes and visible teeth, consistent across frames. The emotion is clearly happy and natural. Preservation is acceptable but not perfect—background elements (e.g., lighting, foliage) show slight changes, possibly due to re-rendering or camera angle adjustment, which may not perfectly match the original. Realism is high: skin texture, lighting, and motion are smooth and natural. Consistency is good: the laugh is stable over time with no temporal drift or artifacts."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 5,\n \"preservation\": 3,\n \"realism\": 4,\n \"consistency\": 4,\n \"reason\": \"The instruction to make the man laugh happily is fully executed: his facial expression shows a genuine, wide smile with crinkled eyes and visible teeth, consistent across frames. The emotion is clearly happy and natural. Preservation is acceptable but not perfect—background elements (e.g., lighting, foliage) show slight changes, possibly due to re-rendering or camera angle adjustment, which may not perfectly match the original. Realism is high: skin texture, lighting, and motion are smooth and natural. Consistency is good: the laugh is stable over time with no temporal drift or artifacts.\"\n}",
+ "receipt_digest": "0f41b7f2be8e95f0c8f52b48f95d98f185127d3ac9942478fdb45cbbb53ad0c1",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 5.0,
+ "accuracy_penalty": 1.0,
+ "consistency": 4.0,
+ "overall_score_paper_0_to_100": 86.0,
+ "overall_score_raw_0_to_5": 4.3,
+ "preservation": 3.0,
+ "realism": 4.0,
+ "reason": "The instruction to make the man laugh happily is fully executed: his facial expression shows a genuine, wide smile with crinkled eyes and visible teeth, consistent across frames. The emotion is clearly happy and natural. Preservation is acceptable but not perfect—background elements (e.g., lighting, foliage) show slight changes, possibly due to re-rendering or camera angle adjustment, which may not perfectly match the original. Realism is high: skin texture, lighting, and motion are smooth and natural. Consistency is good: the laugh is stable over time with no temporal drift or artifacts."
+ },
+ "source_video_sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "subcategory": "Facial_expression_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/reco_ori_wan21_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/reco_ori_wan21_1p3b.json
new file mode 100644
index 0000000..b462ceb
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/reco_ori_wan21_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Semantic",
+ "created_at_utc": "2026-09-14T19:36:46Z",
+ "edited_video_sha256": "a23c0cb3676dc3904957c80dad444a59651862972ecc31155bc01fd40d525246",
+ "editing_instruction": "Make the man laugh happily.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "ce02857d7678e154c3e4ed0e6feab8073b9587fe0c17550586d549a5e1bf13e5",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 25,
+ "item_id": "Te_Facial_expression_change_06",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 3,
+ "preservation": 3,
+ "realism": 2,
+ "reason": "The instruction was to make the man laugh happily, but the edited video replaces the original man with a completely different person (different race, facial features, and hairstyle). This is a fundamental failure in accuracy. The new man does appear to be laughing, but since the subject is not the original man, the edit is incorrect. Background elements are mostly preserved, and motion is somewhat consistent, but the core instruction is violated."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 3,\n \"realism\": 2,\n \"consistency\": 3,\n \"reason\": \"The instruction was to make the man laugh happily, but the edited video replaces the original man with a completely different person (different race, facial features, and hairstyle). This is a fundamental failure in accuracy. The new man does appear to be laughing, but since the subject is not the original man, the edit is incorrect. Background elements are mostly preserved, and motion is somewhat consistent, but the core instruction is violated.\"\n}",
+ "receipt_digest": "a5f11221b5dce4e6213615e7070d0d0a64d0da978f829208dc9aebdffaf002e9",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 3.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 3.0,
+ "realism": 2.0,
+ "reason": "The instruction was to make the man laugh happily, but the edited video replaces the original man with a completely different person (different race, facial features, and hairstyle). This is a fundamental failure in accuracy. The new man does appear to be laughing, but since the subject is not the original man, the edit is incorrect. Background elements are mostly preserved, and motion is somewhat consistent, but the core instruction is violated."
+ },
+ "source_video_sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "subcategory": "Facial_expression_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/vace_wan21_1p3b_fullframe.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/vace_wan21_1p3b_fullframe.json
new file mode 100644
index 0000000..373ee9c
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/vace_wan21_1p3b_fullframe.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Semantic",
+ "created_at_utc": "2026-09-14T19:25:19Z",
+ "edited_video_sha256": "f2194f1c764e6f1f00b1e9e59c0c213e4e574dfa3024cef0d565ba165d081748",
+ "editing_instruction": "Make the man laugh happily.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "dd144add239638ebbc6a7f5dd0112fdc48e4b0eb193048b94cd97eecdeacb179",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 25,
+ "item_id": "Te_Facial_expression_change_06",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 3,
+ "preservation": 3,
+ "realism": 3,
+ "reason": "The instruction was to make the man laugh happily, but the edited video shows no change in facial expression — the man remains serious and neutral throughout. There is no visible laughter, smile, or joyful expression. The edit fails to alter the intended emotion, making accuracy 0. Background elements are mostly preserved, with some color and lighting changes (e.g., brighter green foliage), which is acceptable for preservation. Realism and consistency are decent as the motion and lighting are smooth, but they are irrelevant since the core instruction was not followed."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 3,\n \"realism\": 3,\n \"consistency\": 3,\n \"reason\": \"The instruction was to make the man laugh happily, but the edited video shows no change in facial expression — the man remains serious and neutral throughout. There is no visible laughter, smile, or joyful expression. The edit fails to alter the intended emotion, making accuracy 0. Background elements are mostly preserved, with some color and lighting changes (e.g., brighter green foliage), which is acceptable for preservation. Realism and consistency are decent as the motion and lighting are smooth, but they are irrelevant since the core instruction was not followed.\"\n}",
+ "receipt_digest": "75d01682dd5bed20b44f7c09780394e9e17bfbd94896ec848a088897fbf01ef6",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 3.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 3.0,
+ "realism": 3.0,
+ "reason": "The instruction was to make the man laugh happily, but the edited video shows no change in facial expression — the man remains serious and neutral throughout. There is no visible laughter, smile, or joyful expression. The edit fails to alter the intended emotion, making accuracy 0. Background elements are mostly preserved, with some color and lighting changes (e.g., brighter green foliage), which is acceptable for preservation. Realism and consistency are decent as the motion and lighting are smooth, but they are irrelevant since the core instruction was not followed."
+ },
+ "source_video_sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "subcategory": "Facial_expression_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/zebra.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/zebra.json
new file mode 100644
index 0000000..0372d50
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Facial_expression_change_06/zebra.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Semantic",
+ "created_at_utc": "2026-09-22T09:47:05Z",
+ "edited_video_sha256": "231b466565101154bf9185a812dbffa1195988d8873b746f82d393c913ad66a1",
+ "editing_instruction": "Make the man laugh happily.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "149da52dc20cd8ac0dcb568d1eb13e785bc243ac1c0d1be8003e09e0e9938f7b",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 25,
+ "item_id": "Te_Facial_expression_change_06",
+ "parsed_judge_output": {
+ "accuracy": 5,
+ "consistency": 5,
+ "preservation": 5,
+ "realism": 5,
+ "reason": "The man's expression has been successfully changed to a happy laugh, with natural mouth opening, visible teeth, crinkled eyes, and appropriate facial muscle movement. The edit is temporally consistent, with smooth progression of the laugh across frames. Background elements and lighting remain unchanged, preserving the original context. No artifacts or inconsistencies are visible. The edit fully satisfies the instruction with high realism and temporal stability."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 5,\n \"preservation\": 5,\n \"realism\": 5,\n \"consistency\": 5,\n \"reason\": \"The man's expression has been successfully changed to a happy laugh, with natural mouth opening, visible teeth, crinkled eyes, and appropriate facial muscle movement. The edit is temporally consistent, with smooth progression of the laugh across frames. Background elements and lighting remain unchanged, preserving the original context. No artifacts or inconsistencies are visible. The edit fully satisfies the instruction with high realism and temporal stability.\"\n}",
+ "receipt_digest": "3e632455aa44a76d7cc8a6f9a18d6b12ebe50f2bea30631a5991f6703a724278",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 5.0,
+ "accuracy_penalty": 1.0,
+ "consistency": 5.0,
+ "overall_score_paper_0_to_100": 100.0,
+ "overall_score_raw_0_to_5": 5.0,
+ "preservation": 5.0,
+ "realism": 5.0,
+ "reason": "The man's expression has been successfully changed to a happy laugh, with natural mouth opening, visible teeth, crinkled eyes, and appropriate facial muscle movement. The edit is temporally consistent, with smooth progression of the laugh across frames. Background elements and lighting remain unchanged, preserving the original context. No artifacts or inconsistencies are visible. The edit fully satisfies the instruction with high realism and temporal stability."
+ },
+ "source_video_sha256": "25e2d2fff2da815d34244c47b2074f6b6577c9c2a2d1ee38617eaa64523c2d20",
+ "subcategory": "Facial_expression_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/frozen_bernini.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/frozen_bernini.json
new file mode 100644
index 0000000..792ccd5
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/frozen_bernini.json
@@ -0,0 +1,46 @@
+{
+ "category": "Temporal Composition",
+ "created_at_utc": "2026-09-14T19:11:06Z",
+ "edited_video_sha256": "cd28f95da224595c3b323af7042fe788a08cb033637e3b958144287dd92a241e",
+ "editing_instruction": "Delete the action of the girl giving a flying kiss, while maintaining the continuity of the video naturally.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "35bd76d4ea58ddf315ea3630265a781fca041769aa67c81f839b44de7a7e9141",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 180,
+ "item_id": "Te_Motion_deletion_01",
+ "parsed_judge_output": {
+ "accuracy": 3,
+ "consistency": 4,
+ "preservation": 4,
+ "realism": 3,
+ "reason": "The flying kiss action (hand gesture near mouth) has been removed, which satisfies the core instruction. However, the transition is not natural: the girl's expression and head movement become stiff and unnatural after the removal, especially around 0:03-0:05, where her gaze shifts abruptly and her posture appears frozen. This breaks continuity. The background and unrelated elements (lamp, sofa, laptop) are preserved well. Motion is mostly smooth, but the edited segments show slight artifacts and unnatural facial expressions, reducing realism. Temporal consistency is mostly maintained, but the abrupt change in behavior is noticeable."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 3,\n \"preservation\": 4,\n \"realism\": 3,\n \"consistency\": 4,\n \"reason\": \"The flying kiss action (hand gesture near mouth) has been removed, which satisfies the core instruction. However, the transition is not natural: the girl's expression and head movement become stiff and unnatural after the removal, especially around 0:03-0:05, where her gaze shifts abruptly and her posture appears frozen. This breaks continuity. The background and unrelated elements (lamp, sofa, laptop) are preserved well. Motion is mostly smooth, but the edited segments show slight artifacts and unnatural facial expressions, reducing realism. Temporal consistency is mostly maintained, but the abrupt change in behavior is noticeable.\"\n}",
+ "receipt_digest": "49945bbde366671d5716e6292bfc56a89061d96c699e93638fd8e9798b112c21",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 3.0,
+ "accuracy_penalty": 0.6000000000000001,
+ "consistency": 4.0,
+ "overall_score_paper_0_to_100": 52.2,
+ "overall_score_raw_0_to_5": 2.61,
+ "preservation": 4.0,
+ "realism": 3.0,
+ "reason": "The flying kiss action (hand gesture near mouth) has been removed, which satisfies the core instruction. However, the transition is not natural: the girl's expression and head movement become stiff and unnatural after the removal, especially around 0:03-0:05, where her gaze shifts abruptly and her posture appears frozen. This breaks continuity. The background and unrelated elements (lamp, sofa, laptop) are preserved well. Motion is mostly smooth, but the edited segments show slight artifacts and unnatural facial expressions, reducing realism. Temporal consistency is mostly maintained, but the abrupt change in behavior is noticeable."
+ },
+ "source_video_sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "subcategory": "Motion_deletion"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/kiwi_edit_5b_instruct_only.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/kiwi_edit_5b_instruct_only.json
new file mode 100644
index 0000000..d06e61b
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/kiwi_edit_5b_instruct_only.json
@@ -0,0 +1,46 @@
+{
+ "category": "Temporal Composition",
+ "created_at_utc": "2026-09-14T21:04:57Z",
+ "edited_video_sha256": "365a6e85a43a125865a6c94e892c32490234771de2cea0874c4ba9a7db803d50",
+ "editing_instruction": "Delete the action of the girl giving a flying kiss, while maintaining the continuity of the video naturally.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "0329899b19a0300bc281bd54a128d9314004f9ffe4eadff295b2fee6ed5195aa",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 180,
+ "item_id": "Te_Motion_deletion_01",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 5,
+ "preservation": 5,
+ "realism": 5,
+ "reason": "The instruction was to delete the action of the girl giving a flying kiss (which occurs around 3.8-5.4 seconds in the original video). However, in the edited video, this action is still present — the girl brings her hand to her mouth and blows a kiss. Therefore, the edit failed to remove the specified action, making accuracy 0. Preservation, realism, and consistency are high as the rest of the video remains unchanged and visually smooth."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 5,\n \"realism\": 5,\n \"consistency\": 5,\n \"reason\": \"The instruction was to delete the action of the girl giving a flying kiss (which occurs around 3.8-5.4 seconds in the original video). However, in the edited video, this action is still present — the girl brings her hand to her mouth and blows a kiss. Therefore, the edit failed to remove the specified action, making accuracy 0. Preservation, realism, and consistency are high as the rest of the video remains unchanged and visually smooth.\"\n}",
+ "receipt_digest": "333d0f0193072f0c18b34e1fa3c6300b2f2afe000ebb4022dd400947acc3e1e9",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 5.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 5.0,
+ "realism": 5.0,
+ "reason": "The instruction was to delete the action of the girl giving a flying kiss (which occurs around 3.8-5.4 seconds in the original video). However, in the edited video, this action is still present — the girl brings her hand to her mouth and blows a kiss. Therefore, the edit failed to remove the specified action, making accuracy 0. Preservation, realism, and consistency are high as the rest of the video remains unchanged and visually smooth."
+ },
+ "source_video_sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "subcategory": "Motion_deletion"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/liveedit_wan21_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/liveedit_wan21_1p3b.json
new file mode 100644
index 0000000..f1a2ba6
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/liveedit_wan21_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Temporal Composition",
+ "created_at_utc": "2026-09-14T21:04:39Z",
+ "edited_video_sha256": "b6758782a0ed97a2aae04121b61ad61fe69b481bae4a9220f4be2aeae381a889",
+ "editing_instruction": "Delete the action of the girl giving a flying kiss, while maintaining the continuity of the video naturally.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "fc731a49e6c3d35b5945782f04ebc463488a0e3e5ede54e972425835c208e5d7",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 180,
+ "item_id": "Te_Motion_deletion_01",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 2,
+ "preservation": 2,
+ "realism": 2,
+ "reason": "The instruction was to delete the action of the girl giving a flying kiss, but the edited video not only retains the flying kiss gesture but also adds unrealistic red wings behind her, which is a complete deviation from the original instruction. The action is not removed; instead, it is enhanced with fantasy elements. This is a major failure in accuracy. Preservation is low because the background and lighting are slightly altered (e.g., lamp brightness, wall texture), and the wings are a new element. Realism is poor due to the added wings and unnatural motion. Consistency is compromised as the wings appear and remain throughout, creating a temporal drift."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 2,\n \"consistency\": 2,\n \"reason\": \"The instruction was to delete the action of the girl giving a flying kiss, but the edited video not only retains the flying kiss gesture but also adds unrealistic red wings behind her, which is a complete deviation from the original instruction. The action is not removed; instead, it is enhanced with fantasy elements. This is a major failure in accuracy. Preservation is low because the background and lighting are slightly altered (e.g., lamp brightness, wall texture), and the wings are a new element. Realism is poor due to the added wings and unnatural motion. Consistency is compromised as the wings appear and remain throughout, creating a temporal drift.\"\n}",
+ "receipt_digest": "f4f8ed287a1d8763a7f2653ad55ebfe0de953a32affc6d5558d91b0e0c7a96c9",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 2.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 2.0,
+ "reason": "The instruction was to delete the action of the girl giving a flying kiss, but the edited video not only retains the flying kiss gesture but also adds unrealistic red wings behind her, which is a complete deviation from the original instruction. The action is not removed; instead, it is enhanced with fantasy elements. This is a major failure in accuracy. Preservation is low because the background and lighting are slightly altered (e.g., lamp brightness, wall texture), and the wings are a new element. Realism is poor due to the added wings and unnatural motion. Consistency is compromised as the wings appear and remain throughout, creating a temporal drift."
+ },
+ "source_video_sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "subcategory": "Motion_deletion"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/lucy_edit_1p1_dev_5b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/lucy_edit_1p1_dev_5b.json
new file mode 100644
index 0000000..8dde476
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/lucy_edit_1p1_dev_5b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Temporal Composition",
+ "created_at_utc": "2026-09-14T19:07:41Z",
+ "edited_video_sha256": "bacf48ee36c1b5fad294c9ea3a3f6376792946c23297800b67f197ab3bd1d0be",
+ "editing_instruction": "Delete the action of the girl giving a flying kiss, while maintaining the continuity of the video naturally.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "958406a72abc3d1d3ccff777bbe657b4e77061cb2ecce850fc4a0f8546716c3a",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 180,
+ "item_id": "Te_Motion_deletion_01",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 5,
+ "preservation": 5,
+ "realism": 5,
+ "reason": "The instruction was to delete the action of the girl giving a flying kiss (which occurs around 3.8-5.4 seconds in the original video). However, in the edited video, this action is still present — the girl brings her hand to her lips and blows a kiss. Therefore, the edit failed to remove the specified action, making accuracy 0. All other aspects (background, lighting, motion, continuity) are preserved and realistic, but accuracy is the top priority and must be low due to the failure to follow the instruction."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 5,\n \"realism\": 5,\n \"consistency\": 5,\n \"reason\": \"The instruction was to delete the action of the girl giving a flying kiss (which occurs around 3.8-5.4 seconds in the original video). However, in the edited video, this action is still present — the girl brings her hand to her lips and blows a kiss. Therefore, the edit failed to remove the specified action, making accuracy 0. All other aspects (background, lighting, motion, continuity) are preserved and realistic, but accuracy is the top priority and must be low due to the failure to follow the instruction.\"\n}",
+ "receipt_digest": "48a335e433dc03534f99106a1099d65cde0f0392b4aaf812c38fcdc86d90bded",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 5.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 5.0,
+ "realism": 5.0,
+ "reason": "The instruction was to delete the action of the girl giving a flying kiss (which occurs around 3.8-5.4 seconds in the original video). However, in the edited video, this action is still present — the girl brings her hand to her lips and blows a kiss. Therefore, the edit failed to remove the specified action, making accuracy 0. All other aspects (background, lighting, motion, continuity) are preserved and realistic, but accuracy is the top priority and must be low due to the failure to follow the instruction."
+ },
+ "source_video_sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "subcategory": "Motion_deletion"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/omnivideo2_official_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/omnivideo2_official_1p3b.json
new file mode 100644
index 0000000..dbccdee
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/omnivideo2_official_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Temporal Composition",
+ "created_at_utc": "2026-09-14T19:15:12Z",
+ "edited_video_sha256": "362e2f0a8b72ccdac9fc6bb89eca33b5f54333d4b5454e7ba8c6ecf249aaa699",
+ "editing_instruction": "Delete the action of the girl giving a flying kiss, while maintaining the continuity of the video naturally.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "1b79bbf410c6adf53abd87332bc0cc6930b4f54494ede953234b3355e38afefb",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 180,
+ "item_id": "Te_Motion_deletion_01",
+ "parsed_judge_output": {
+ "accuracy": 2,
+ "consistency": 3,
+ "preservation": 3,
+ "realism": 3,
+ "reason": "The flying kiss action (hand moving toward mouth and blowing a kiss) is not fully removed — it is still present in the edited video, particularly around 0:03-0:05. The instruction was to delete this action, but it remains, making accuracy low. The hand motion appears slightly unnatural and inconsistent with the original, suggesting editing artifacts. Preservation is acceptable as background elements (lamp, couch, wall) remain unchanged. Realism and consistency are moderate — motion is smooth but the hand gesture is not fully corrected, and there is slight temporal drift in the hand’s position across frames."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 2,\n \"preservation\": 3,\n \"realism\": 3,\n \"consistency\": 3,\n \"reason\": \"The flying kiss action (hand moving toward mouth and blowing a kiss) is not fully removed — it is still present in the edited video, particularly around 0:03-0:05. The instruction was to delete this action, but it remains, making accuracy low. The hand motion appears slightly unnatural and inconsistent with the original, suggesting editing artifacts. Preservation is acceptable as background elements (lamp, couch, wall) remain unchanged. Realism and consistency are moderate — motion is smooth but the hand gesture is not fully corrected, and there is slight temporal drift in the hand’s position across frames.\"\n}",
+ "receipt_digest": "5a48fac5ef73ed38b7a25fa235c5ae338afe4d14adf160ad8da1a5d4ce51a139",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 2.0,
+ "accuracy_penalty": 0.4,
+ "consistency": 3.0,
+ "overall_score_paper_0_to_100": 32.0,
+ "overall_score_raw_0_to_5": 1.6,
+ "preservation": 3.0,
+ "realism": 3.0,
+ "reason": "The flying kiss action (hand moving toward mouth and blowing a kiss) is not fully removed — it is still present in the edited video, particularly around 0:03-0:05. The instruction was to delete this action, but it remains, making accuracy low. The hand motion appears slightly unnatural and inconsistent with the original, suggesting editing artifacts. Preservation is acceptable as background elements (lamp, couch, wall) remain unchanged. Realism and consistency are moderate — motion is smooth but the hand gesture is not fully corrected, and there is slight temporal drift in the hand’s position across frames."
+ },
+ "source_video_sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "subcategory": "Motion_deletion"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/reco_ori_wan21_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/reco_ori_wan21_1p3b.json
new file mode 100644
index 0000000..e477812
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/reco_ori_wan21_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Temporal Composition",
+ "created_at_utc": "2026-09-14T20:58:05Z",
+ "edited_video_sha256": "8febc79000879f423a5236cff8d712953999a152911f8d606c50bff0d458920f",
+ "editing_instruction": "Delete the action of the girl giving a flying kiss, while maintaining the continuity of the video naturally.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "ce02857d7678e154c3e4ed0e6feab8073b9587fe0c17550586d549a5e1bf13e5",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 180,
+ "item_id": "Te_Motion_deletion_01",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 2,
+ "preservation": 0,
+ "realism": 2,
+ "reason": "The instruction was to delete the action of the girl giving a flying kiss while maintaining natural continuity. However, the flying kiss action (hand near mouth, puckered lips) is still present in the edited video at 0:03-0:05. Additionally, the entire scene has been altered with a completely different color palette (yellow/orange walls, blue laptop, different artwork), which violates preservation of unrelated content. The girl’s appearance (skin tone, hair, clothing) is also changed, indicating a major failure in preserving the original context. The editing is not accurate, and the changes are not subtle or natural. Realism is low due to unnatural skin tones and lighting, and consistency is poor due to abrupt color shifts and inconsistent object appearance."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 0,\n \"realism\": 2,\n \"consistency\": 2,\n \"reason\": \"The instruction was to delete the action of the girl giving a flying kiss while maintaining natural continuity. However, the flying kiss action (hand near mouth, puckered lips) is still present in the edited video at 0:03-0:05. Additionally, the entire scene has been altered with a completely different color palette (yellow/orange walls, blue laptop, different artwork), which violates preservation of unrelated content. The girl’s appearance (skin tone, hair, clothing) is also changed, indicating a major failure in preserving the original context. The editing is not accurate, and the changes are not subtle or natural. Realism is low due to unnatural skin tones and lighting, and consistency is poor due to abrupt color shifts and inconsistent object appearance.\"\n}",
+ "receipt_digest": "05afc9f9c6f83463b232b0e15cbf0088e0558309f208add4eb383ac5933a63cd",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 2.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 0.0,
+ "realism": 2.0,
+ "reason": "The instruction was to delete the action of the girl giving a flying kiss while maintaining natural continuity. However, the flying kiss action (hand near mouth, puckered lips) is still present in the edited video at 0:03-0:05. Additionally, the entire scene has been altered with a completely different color palette (yellow/orange walls, blue laptop, different artwork), which violates preservation of unrelated content. The girl’s appearance (skin tone, hair, clothing) is also changed, indicating a major failure in preserving the original context. The editing is not accurate, and the changes are not subtle or natural. Realism is low due to unnatural skin tones and lighting, and consistency is poor due to abrupt color shifts and inconsistent object appearance."
+ },
+ "source_video_sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "subcategory": "Motion_deletion"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/vace_wan21_1p3b_fullframe.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/vace_wan21_1p3b_fullframe.json
new file mode 100644
index 0000000..16f5ee8
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/vace_wan21_1p3b_fullframe.json
@@ -0,0 +1,46 @@
+{
+ "category": "Temporal Composition",
+ "created_at_utc": "2026-09-14T20:50:16Z",
+ "edited_video_sha256": "c4a5537e65793a14c7f0aa816056105672f6864ff8b4bf89230f08a1ea28e371",
+ "editing_instruction": "Delete the action of the girl giving a flying kiss, while maintaining the continuity of the video naturally.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "dd144add239638ebbc6a7f5dd0112fdc48e4b0eb193048b94cd97eecdeacb179",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 180,
+ "item_id": "Te_Motion_deletion_01",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 2,
+ "preservation": 2,
+ "realism": 2,
+ "reason": "The instruction was to delete the action of the girl giving a flying kiss (which occurs around 3.8-5.4 seconds in the original video). However, in the edited video, the flying kiss action is still present and unchanged. Therefore, the edit failed to remove the specified action, making accuracy 0. Additionally, the girl's hair color has been altered from brown to red, which is an unrelated change violating preservation. The video also exhibits unnatural color shifts and motion artifacts, reducing realism and consistency."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 2,\n \"consistency\": 2,\n \"reason\": \"The instruction was to delete the action of the girl giving a flying kiss (which occurs around 3.8-5.4 seconds in the original video). However, in the edited video, the flying kiss action is still present and unchanged. Therefore, the edit failed to remove the specified action, making accuracy 0. Additionally, the girl's hair color has been altered from brown to red, which is an unrelated change violating preservation. The video also exhibits unnatural color shifts and motion artifacts, reducing realism and consistency.\"\n}",
+ "receipt_digest": "47e54509fba61dce775aaddf16ac99fac9b6381662a0c49818bb52dd3e99ba5f",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 2.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 2.0,
+ "reason": "The instruction was to delete the action of the girl giving a flying kiss (which occurs around 3.8-5.4 seconds in the original video). However, in the edited video, the flying kiss action is still present and unchanged. Therefore, the edit failed to remove the specified action, making accuracy 0. Additionally, the girl's hair color has been altered from brown to red, which is an unrelated change violating preservation. The video also exhibits unnatural color shifts and motion artifacts, reducing realism and consistency."
+ },
+ "source_video_sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "subcategory": "Motion_deletion"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/zebra.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/zebra.json
new file mode 100644
index 0000000..554ee2a
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_deletion_01/zebra.json
@@ -0,0 +1,46 @@
+{
+ "category": "Temporal Composition",
+ "created_at_utc": "2026-09-22T06:18:30Z",
+ "edited_video_sha256": "1d4de93dce10739e46069ec9f269a92de765c2f970dfa85d8de42a1875640ddd",
+ "editing_instruction": "Delete the action of the girl giving a flying kiss, while maintaining the continuity of the video naturally.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "149da52dc20cd8ac0dcb568d1eb13e785bc243ac1c0d1be8003e09e0e9938f7b",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 180,
+ "item_id": "Te_Motion_deletion_01",
+ "parsed_judge_output": {
+ "accuracy": 5,
+ "consistency": 5,
+ "preservation": 5,
+ "realism": 5,
+ "reason": "The flying kiss action (hand gesture near mouth) has been completely and naturally removed. The girl's expression and head movement transition smoothly from waving to looking down at the laptop, maintaining natural continuity. Background elements (lamp, sofa, wall art) remain unchanged. Motion is smooth, no artifacts, and temporal consistency is preserved throughout. The edit is seamless and fully aligns with the instruction."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 5,\n \"preservation\": 5,\n \"realism\": 5,\n \"consistency\": 5,\n \"reason\": \"The flying kiss action (hand gesture near mouth) has been completely and naturally removed. The girl's expression and head movement transition smoothly from waving to looking down at the laptop, maintaining natural continuity. Background elements (lamp, sofa, wall art) remain unchanged. Motion is smooth, no artifacts, and temporal consistency is preserved throughout. The edit is seamless and fully aligns with the instruction.\"\n}",
+ "receipt_digest": "6b2559547f7c811a3503e3acf5161e074a125a65020310fb184394f3c955c6f0",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 5.0,
+ "accuracy_penalty": 1.0,
+ "consistency": 5.0,
+ "overall_score_paper_0_to_100": 100.0,
+ "overall_score_raw_0_to_5": 5.0,
+ "preservation": 5.0,
+ "realism": 5.0,
+ "reason": "The flying kiss action (hand gesture near mouth) has been completely and naturally removed. The girl's expression and head movement transition smoothly from waving to looking down at the laptop, maintaining natural continuity. Background elements (lamp, sofa, wall art) remain unchanged. Motion is smooth, no artifacts, and temporal consistency is preserved throughout. The edit is seamless and fully aligns with the instruction."
+ },
+ "source_video_sha256": "7c51ba4b1218a1cd4f153ab9ae174dce0d47a56f5e6b32b89a885a1d5a8ad6ba",
+ "subcategory": "Motion_deletion"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/frozen_bernini.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/frozen_bernini.json
new file mode 100644
index 0000000..58e8e66
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/frozen_bernini.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T18:19:19Z",
+ "edited_video_sha256": "e166a44a9312219d78537f13f49e8e04800a1814f02290ea39334ff7efddafd9",
+ "editing_instruction": "Reduce the frequency of the fan's rotation on the ceiling.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "35bd76d4ea58ddf315ea3630265a781fca041769aa67c81f839b44de7a7e9141",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 81,
+ "item_id": "Te_Motion_frequency_change_02",
+ "parsed_judge_output": {
+ "accuracy": 2,
+ "consistency": 4,
+ "preservation": 5,
+ "realism": 3,
+ "reason": "The fan's rotation frequency was not meaningfully reduced; the blades still appear to spin at a similar high speed as in the original video, with no noticeable decrease in RPM. The instruction to reduce frequency is not correctly followed. However, unrelated elements (furniture, lighting, camera angle) are preserved perfectly. Motion appears somewhat realistic but with slight blur artifacts. Temporal consistency is mostly maintained, though the fan's motion is inconsistent with the intended edit."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 2,\n \"preservation\": 5,\n \"realism\": 3,\n \"consistency\": 4,\n \"reason\": \"The fan's rotation frequency was not meaningfully reduced; the blades still appear to spin at a similar high speed as in the original video, with no noticeable decrease in RPM. The instruction to reduce frequency is not correctly followed. However, unrelated elements (furniture, lighting, camera angle) are preserved perfectly. Motion appears somewhat realistic but with slight blur artifacts. Temporal consistency is mostly maintained, though the fan's motion is inconsistent with the intended edit.\"\n}",
+ "receipt_digest": "74e3b830b2ea6700c198c6b8053f5e32551476030fc02440746ce07a7c04da9f",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 2.0,
+ "accuracy_penalty": 0.4,
+ "consistency": 4.0,
+ "overall_score_paper_0_to_100": 36.4,
+ "overall_score_raw_0_to_5": 1.82,
+ "preservation": 5.0,
+ "realism": 3.0,
+ "reason": "The fan's rotation frequency was not meaningfully reduced; the blades still appear to spin at a similar high speed as in the original video, with no noticeable decrease in RPM. The instruction to reduce frequency is not correctly followed. However, unrelated elements (furniture, lighting, camera angle) are preserved perfectly. Motion appears somewhat realistic but with slight blur artifacts. Temporal consistency is mostly maintained, though the fan's motion is inconsistent with the intended edit."
+ },
+ "source_video_sha256": "eee3a86a3441fb2617dcc8323c157d39fc83dc2bd2c0edf1ea860cd6d9c903d1",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/kiwi_edit_5b_instruct_only.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/kiwi_edit_5b_instruct_only.json
new file mode 100644
index 0000000..a89daa4
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/kiwi_edit_5b_instruct_only.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T20:12:08Z",
+ "edited_video_sha256": "9fc6f9989ff621d63f3288406708eed133a514e7a5f5668eea57832e82ada495",
+ "editing_instruction": "Reduce the frequency of the fan's rotation on the ceiling.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "0329899b19a0300bc281bd54a128d9314004f9ffe4eadff295b2fee6ed5195aa",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 81,
+ "item_id": "Te_Motion_frequency_change_02",
+ "parsed_judge_output": {
+ "accuracy": 2,
+ "consistency": 3,
+ "preservation": 4,
+ "realism": 3,
+ "reason": "The fan's rotation frequency appears reduced, but the motion blur is inconsistent and unnatural—sometimes the blades are almost static, other times they blur excessively, suggesting incorrect temporal interpolation. The fan's center also appears distorted or replaced with a white, glowing object in some frames, which is not present in the original. This indicates a failure in accurately modifying only the rotation speed. Unrelated elements (furniture, lighting, background) are preserved well, and motion is mostly smooth, but the fan's visual artifacts break realism and consistency."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 2,\n \"preservation\": 4,\n \"realism\": 3,\n \"consistency\": 3,\n \"reason\": \"The fan's rotation frequency appears reduced, but the motion blur is inconsistent and unnatural—sometimes the blades are almost static, other times they blur excessively, suggesting incorrect temporal interpolation. The fan's center also appears distorted or replaced with a white, glowing object in some frames, which is not present in the original. This indicates a failure in accurately modifying only the rotation speed. Unrelated elements (furniture, lighting, background) are preserved well, and motion is mostly smooth, but the fan's visual artifacts break realism and consistency.\"\n}",
+ "receipt_digest": "fd511d9785aee7e30203e59c93c7030e297050020241fd25fe3bba33adf8cedb",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 2.0,
+ "accuracy_penalty": 0.4,
+ "consistency": 3.0,
+ "overall_score_paper_0_to_100": 33.6,
+ "overall_score_raw_0_to_5": 1.68,
+ "preservation": 4.0,
+ "realism": 3.0,
+ "reason": "The fan's rotation frequency appears reduced, but the motion blur is inconsistent and unnatural—sometimes the blades are almost static, other times they blur excessively, suggesting incorrect temporal interpolation. The fan's center also appears distorted or replaced with a white, glowing object in some frames, which is not present in the original. This indicates a failure in accurately modifying only the rotation speed. Unrelated elements (furniture, lighting, background) are preserved well, and motion is mostly smooth, but the fan's visual artifacts break realism and consistency."
+ },
+ "source_video_sha256": "eee3a86a3441fb2617dcc8323c157d39fc83dc2bd2c0edf1ea860cd6d9c903d1",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/liveedit_wan21_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/liveedit_wan21_1p3b.json
new file mode 100644
index 0000000..53108a8
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/liveedit_wan21_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T20:09:05Z",
+ "edited_video_sha256": "a37e2101b84e691a17865a185c5900287036f1443d241d929a0fbe8c65180080",
+ "editing_instruction": "Reduce the frequency of the fan's rotation on the ceiling.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "fc731a49e6c3d35b5945782f04ebc463488a0e3e5ede54e972425835c208e5d7",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 81,
+ "item_id": "Te_Motion_frequency_change_02",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 2,
+ "preservation": 2,
+ "realism": 2,
+ "reason": "The instruction was to reduce the frequency (i.e., slow down) the fan's rotation. However, the edited video shows a completely different fan model with a circular black ring and a different blade design, which is not present in the original video. This is a major accuracy failure. Additionally, the fan's motion appears unnatural and inconsistent, with visible artifacts and blurring. The rest of the room is mostly preserved, but there are minor distortions in lighting and object edges, indicating poor preservation. Realism and consistency are compromised due to the unrealistic fan design and motion."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 2,\n \"consistency\": 2,\n \"reason\": \"The instruction was to reduce the frequency (i.e., slow down) the fan's rotation. However, the edited video shows a completely different fan model with a circular black ring and a different blade design, which is not present in the original video. This is a major accuracy failure. Additionally, the fan's motion appears unnatural and inconsistent, with visible artifacts and blurring. The rest of the room is mostly preserved, but there are minor distortions in lighting and object edges, indicating poor preservation. Realism and consistency are compromised due to the unrealistic fan design and motion.\"\n}",
+ "receipt_digest": "dff18bdd291f8552ecf01e93f300aaab75ca44b10b53f1e10e5c5d57d59f7521",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 2.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 2.0,
+ "reason": "The instruction was to reduce the frequency (i.e., slow down) the fan's rotation. However, the edited video shows a completely different fan model with a circular black ring and a different blade design, which is not present in the original video. This is a major accuracy failure. Additionally, the fan's motion appears unnatural and inconsistent, with visible artifacts and blurring. The rest of the room is mostly preserved, but there are minor distortions in lighting and object edges, indicating poor preservation. Realism and consistency are compromised due to the unrealistic fan design and motion."
+ },
+ "source_video_sha256": "eee3a86a3441fb2617dcc8323c157d39fc83dc2bd2c0edf1ea860cd6d9c903d1",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/lucy_edit_1p1_dev_5b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/lucy_edit_1p1_dev_5b.json
new file mode 100644
index 0000000..fb7722f
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/lucy_edit_1p1_dev_5b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T18:17:15Z",
+ "edited_video_sha256": "7faa97ec69c6559f1251d072e506a877e8e85f52239893bebd3a33d66b6269db",
+ "editing_instruction": "Reduce the frequency of the fan's rotation on the ceiling.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "958406a72abc3d1d3ccff777bbe657b4e77061cb2ecce850fc4a0f8546716c3a",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 81,
+ "item_id": "Te_Motion_frequency_change_02",
+ "parsed_judge_output": {
+ "accuracy": 4,
+ "consistency": 5,
+ "preservation": 5,
+ "realism": 4,
+ "reason": "The fan's rotation frequency is visibly reduced compared to the original video, fulfilling the instruction. The motion appears slower and more deliberate, with blades showing more distinct positions over time. No other elements in the scene (furniture, lighting, camera angle) are altered, preserving the environment. Motion is smooth and realistic with no noticeable artifacts. Temporal consistency is maintained across frames. Minor imperfection: the fan's motion blur is slightly inconsistent in some frames, but overall, the edit is accurate and well-executed."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 4,\n \"preservation\": 5,\n \"realism\": 4,\n \"consistency\": 5,\n \"reason\": \"The fan's rotation frequency is visibly reduced compared to the original video, fulfilling the instruction. The motion appears slower and more deliberate, with blades showing more distinct positions over time. No other elements in the scene (furniture, lighting, camera angle) are altered, preserving the environment. Motion is smooth and realistic with no noticeable artifacts. Temporal consistency is maintained across frames. Minor imperfection: the fan's motion blur is slightly inconsistent in some frames, but overall, the edit is accurate and well-executed.\"\n}",
+ "receipt_digest": "e1b4a86819bf49f69fdbf3c6abccd8873a577e1f5d464f488bef404c97d91816",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 4.0,
+ "accuracy_penalty": 0.8,
+ "consistency": 5.0,
+ "overall_score_paper_0_to_100": 77.6,
+ "overall_score_raw_0_to_5": 3.88,
+ "preservation": 5.0,
+ "realism": 4.0,
+ "reason": "The fan's rotation frequency is visibly reduced compared to the original video, fulfilling the instruction. The motion appears slower and more deliberate, with blades showing more distinct positions over time. No other elements in the scene (furniture, lighting, camera angle) are altered, preserving the environment. Motion is smooth and realistic with no noticeable artifacts. Temporal consistency is maintained across frames. Minor imperfection: the fan's motion blur is slightly inconsistent in some frames, but overall, the edit is accurate and well-executed."
+ },
+ "source_video_sha256": "eee3a86a3441fb2617dcc8323c157d39fc83dc2bd2c0edf1ea860cd6d9c903d1",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/omnivideo2_official_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/omnivideo2_official_1p3b.json
new file mode 100644
index 0000000..0f9ca8a
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/omnivideo2_official_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T18:21:32Z",
+ "edited_video_sha256": "d030b29c60403f5c51b9c8d9272eb81853e270f81199ec05918b1d9db971dc78",
+ "editing_instruction": "Reduce the frequency of the fan's rotation on the ceiling.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "1b79bbf410c6adf53abd87332bc0cc6930b4f54494ede953234b3355e38afefb",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 81,
+ "item_id": "Te_Motion_frequency_change_02",
+ "parsed_judge_output": {
+ "accuracy": 2,
+ "consistency": 2,
+ "preservation": 1,
+ "realism": 3,
+ "reason": "The fan's rotation frequency appears slightly reduced, but the change is minimal and not clearly accurate. More critically, the edited video shows significant preservation errors: the coffee table decor (wooden beads and bowl) is replaced with a textured basket, and the patterned throw pillow is altered. The dining chairs and kitchen area also show minor structural changes. These unrelated changes violate the preservation rule. Temporal consistency is compromised due to these artifacts and inconsistencies in object appearance across frames. Realism is acceptable but not high due to the unnatural changes in static objects."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 2,\n \"preservation\": 1,\n \"realism\": 3,\n \"consistency\": 2,\n \"reason\": \"The fan's rotation frequency appears slightly reduced, but the change is minimal and not clearly accurate. More critically, the edited video shows significant preservation errors: the coffee table decor (wooden beads and bowl) is replaced with a textured basket, and the patterned throw pillow is altered. The dining chairs and kitchen area also show minor structural changes. These unrelated changes violate the preservation rule. Temporal consistency is compromised due to these artifacts and inconsistencies in object appearance across frames. Realism is acceptable but not high due to the unnatural changes in static objects.\"\n}",
+ "receipt_digest": "554235751614b669c89c0ae2a3e37d0d6f6efb8cd67ce558ba224b693a669e9f",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 2.0,
+ "accuracy_penalty": 0.4,
+ "consistency": 2.0,
+ "overall_score_paper_0_to_100": 27.6,
+ "overall_score_raw_0_to_5": 1.38,
+ "preservation": 1.0,
+ "realism": 3.0,
+ "reason": "The fan's rotation frequency appears slightly reduced, but the change is minimal and not clearly accurate. More critically, the edited video shows significant preservation errors: the coffee table decor (wooden beads and bowl) is replaced with a textured basket, and the patterned throw pillow is altered. The dining chairs and kitchen area also show minor structural changes. These unrelated changes violate the preservation rule. Temporal consistency is compromised due to these artifacts and inconsistencies in object appearance across frames. Realism is acceptable but not high due to the unnatural changes in static objects."
+ },
+ "source_video_sha256": "eee3a86a3441fb2617dcc8323c157d39fc83dc2bd2c0edf1ea860cd6d9c903d1",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/reco_ori_wan21_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/reco_ori_wan21_1p3b.json
new file mode 100644
index 0000000..22fee0b
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/reco_ori_wan21_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T20:05:59Z",
+ "edited_video_sha256": "90db4129783fa802f016a01e0bd78b02327445d4a423c3fb499b103d2b75c21d",
+ "editing_instruction": "Reduce the frequency of the fan's rotation on the ceiling.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "ce02857d7678e154c3e4ed0e6feab8073b9587fe0c17550586d549a5e1bf13e5",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 81,
+ "item_id": "Te_Motion_frequency_change_02",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 0,
+ "preservation": 0,
+ "realism": 0,
+ "reason": "The edited video is a split-screen with a distorted, non-functional representation of the room. The ceiling fan is not visible in the left half, and the right half shows only partial, inconsistent frames. The fan's rotation frequency is not reduced — it is either absent or not properly rendered. The entire scene is corrupted, with objects misaligned and lighting inconsistent. This is a complete failure to follow the instruction and preserve the original content."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 0,\n \"realism\": 0,\n \"consistency\": 0,\n \"reason\": \"The edited video is a split-screen with a distorted, non-functional representation of the room. The ceiling fan is not visible in the left half, and the right half shows only partial, inconsistent frames. The fan's rotation frequency is not reduced — it is either absent or not properly rendered. The entire scene is corrupted, with objects misaligned and lighting inconsistent. This is a complete failure to follow the instruction and preserve the original content.\"\n}",
+ "receipt_digest": "e2b92f76ef2faf80b2d490a62f8f8aef52be6812694040a9d551933d53508909",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 0.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 0.0,
+ "realism": 0.0,
+ "reason": "The edited video is a split-screen with a distorted, non-functional representation of the room. The ceiling fan is not visible in the left half, and the right half shows only partial, inconsistent frames. The fan's rotation frequency is not reduced — it is either absent or not properly rendered. The entire scene is corrupted, with objects misaligned and lighting inconsistent. This is a complete failure to follow the instruction and preserve the original content."
+ },
+ "source_video_sha256": "eee3a86a3441fb2617dcc8323c157d39fc83dc2bd2c0edf1ea860cd6d9c903d1",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/vace_wan21_1p3b_fullframe.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/vace_wan21_1p3b_fullframe.json
new file mode 100644
index 0000000..a527f5e
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/vace_wan21_1p3b_fullframe.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T19:56:53Z",
+ "edited_video_sha256": "4fc6bd5fdcedb2703a879b09cc0460dc9da0646ab8d9774fa4ba828f46a4e972",
+ "editing_instruction": "Reduce the frequency of the fan's rotation on the ceiling.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "dd144add239638ebbc6a7f5dd0112fdc48e4b0eb193048b94cd97eecdeacb179",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 81,
+ "item_id": "Te_Motion_frequency_change_02",
+ "parsed_judge_output": {
+ "accuracy": 2,
+ "consistency": 3,
+ "preservation": 2,
+ "realism": 3,
+ "reason": "The fan's rotation frequency appears reduced, but the motion blur is inconsistent and unnatural, suggesting incorrect temporal interpolation. More critically, the coffee table decor has been changed (wooden beads replaced with green and red beads, bowl color altered), which violates preservation rules. The lighting in the kitchen also appears warmer and altered, indicating unintended edits. These changes are not part of the instruction and degrade accuracy and preservation. The fan motion is not perfectly smooth, and temporal consistency is slightly off due to inconsistent blur patterns."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 2,\n \"preservation\": 2,\n \"realism\": 3,\n \"consistency\": 3,\n \"reason\": \"The fan's rotation frequency appears reduced, but the motion blur is inconsistent and unnatural, suggesting incorrect temporal interpolation. More critically, the coffee table decor has been changed (wooden beads replaced with green and red beads, bowl color altered), which violates preservation rules. The lighting in the kitchen also appears warmer and altered, indicating unintended edits. These changes are not part of the instruction and degrade accuracy and preservation. The fan motion is not perfectly smooth, and temporal consistency is slightly off due to inconsistent blur patterns.\"\n}",
+ "receipt_digest": "e28d1874037a011f7f31b7577f491f4277bc9787d9257073b710322002bfba9d",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 2.0,
+ "accuracy_penalty": 0.4,
+ "consistency": 3.0,
+ "overall_score_paper_0_to_100": 30.4,
+ "overall_score_raw_0_to_5": 1.52,
+ "preservation": 2.0,
+ "realism": 3.0,
+ "reason": "The fan's rotation frequency appears reduced, but the motion blur is inconsistent and unnatural, suggesting incorrect temporal interpolation. More critically, the coffee table decor has been changed (wooden beads replaced with green and red beads, bowl color altered), which violates preservation rules. The lighting in the kitchen also appears warmer and altered, indicating unintended edits. These changes are not part of the instruction and degrade accuracy and preservation. The fan motion is not perfectly smooth, and temporal consistency is slightly off due to inconsistent blur patterns."
+ },
+ "source_video_sha256": "eee3a86a3441fb2617dcc8323c157d39fc83dc2bd2c0edf1ea860cd6d9c903d1",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/zebra.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/zebra.json
new file mode 100644
index 0000000..fe1ef50
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_02/zebra.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-22T10:01:07Z",
+ "edited_video_sha256": "20c872bafe12b8de1073240051242732936beb02b3a2ea716fa18a5194ff70de",
+ "editing_instruction": "Reduce the frequency of the fan's rotation on the ceiling.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "149da52dc20cd8ac0dcb568d1eb13e785bc243ac1c0d1be8003e09e0e9938f7b",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 81,
+ "item_id": "Te_Motion_frequency_change_02",
+ "parsed_judge_output": {
+ "accuracy": 5,
+ "consistency": 5,
+ "preservation": 5,
+ "realism": 5,
+ "reason": "The fan's rotation frequency is clearly reduced in the edited video compared to the original, with blades appearing slower and more distinct in motion. The rest of the scene (furniture, lighting, camera angle) remains unchanged, preserving all unrelated elements. Motion is smooth and natural, with no temporal drift or artifacts. The edit is temporally consistent and visually realistic."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 5,\n \"preservation\": 5,\n \"realism\": 5,\n \"consistency\": 5,\n \"reason\": \"The fan's rotation frequency is clearly reduced in the edited video compared to the original, with blades appearing slower and more distinct in motion. The rest of the scene (furniture, lighting, camera angle) remains unchanged, preserving all unrelated elements. Motion is smooth and natural, with no temporal drift or artifacts. The edit is temporally consistent and visually realistic.\"\n}",
+ "receipt_digest": "1c6623e95ebcd49345dcce9580c7b43dd763fc7d0bc27e8f43b2b1447b2e3dba",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 5.0,
+ "accuracy_penalty": 1.0,
+ "consistency": 5.0,
+ "overall_score_paper_0_to_100": 100.0,
+ "overall_score_raw_0_to_5": 5.0,
+ "preservation": 5.0,
+ "realism": 5.0,
+ "reason": "The fan's rotation frequency is clearly reduced in the edited video compared to the original, with blades appearing slower and more distinct in motion. The rest of the scene (furniture, lighting, camera angle) remains unchanged, preserving all unrelated elements. Motion is smooth and natural, with no temporal drift or artifacts. The edit is temporally consistent and visually realistic."
+ },
+ "source_video_sha256": "eee3a86a3441fb2617dcc8323c157d39fc83dc2bd2c0edf1ea860cd6d9c903d1",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/frozen_bernini.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/frozen_bernini.json
new file mode 100644
index 0000000..55af75b
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/frozen_bernini.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T18:21:12Z",
+ "edited_video_sha256": "197736d1ded8d96cfe65d3fd3a8982b9866a60a19d6b5799bfd337258db95700",
+ "editing_instruction": "Slow down the girl's kicking motion.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "35bd76d4ea58ddf315ea3630265a781fca041769aa67c81f839b44de7a7e9141",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 85,
+ "item_id": "Te_Motion_frequency_change_06",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 1,
+ "preservation": 2,
+ "realism": 1,
+ "reason": "The instruction was to slow down the girl's kicking motion, but in the edited video, the girl does not perform any kicks at all. Instead, she stands still or moves minimally, which is a complete failure to follow the instruction. The motion is not slowed — it is removed entirely. This is a major accuracy failure. Preservation is partially maintained in background elements, but some background people and objects (like the TV and tripod) appear in the edited video that were not present in the original, indicating changes to unrelated regions. Realism and consistency are low due to unnatural stillness and temporal drift in the background."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 1,\n \"consistency\": 1,\n \"reason\": \"The instruction was to slow down the girl's kicking motion, but in the edited video, the girl does not perform any kicks at all. Instead, she stands still or moves minimally, which is a complete failure to follow the instruction. The motion is not slowed — it is removed entirely. This is a major accuracy failure. Preservation is partially maintained in background elements, but some background people and objects (like the TV and tripod) appear in the edited video that were not present in the original, indicating changes to unrelated regions. Realism and consistency are low due to unnatural stillness and temporal drift in the background.\"\n}",
+ "receipt_digest": "3b4731b8ae2a637c4d7683c574ad7640a307f937a3ef288cd06abfe0d7499897",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 1.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 1.0,
+ "reason": "The instruction was to slow down the girl's kicking motion, but in the edited video, the girl does not perform any kicks at all. Instead, she stands still or moves minimally, which is a complete failure to follow the instruction. The motion is not slowed — it is removed entirely. This is a major accuracy failure. Preservation is partially maintained in background elements, but some background people and objects (like the TV and tripod) appear in the edited video that were not present in the original, indicating changes to unrelated regions. Realism and consistency are low due to unnatural stillness and temporal drift in the background."
+ },
+ "source_video_sha256": "89033eb8d081cf564f28ff41111f25fcd6bd7f41604c0f4d3e54bb6f950f5ee8",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/kiwi_edit_5b_instruct_only.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/kiwi_edit_5b_instruct_only.json
new file mode 100644
index 0000000..f6f0040
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/kiwi_edit_5b_instruct_only.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T20:14:06Z",
+ "edited_video_sha256": "69143741503ec5c260fb5682fba38a9ac0a8d95a23689f582828454821c4c746",
+ "editing_instruction": "Slow down the girl's kicking motion.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "0329899b19a0300bc281bd54a128d9314004f9ffe4eadff295b2fee6ed5195aa",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 85,
+ "item_id": "Te_Motion_frequency_change_06",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 1,
+ "preservation": 2,
+ "realism": 1,
+ "reason": "The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks are actually faster and more blurred, indicating motion acceleration rather than slowing down. This is a critical accuracy failure. Additionally, background elements like the fan and signage show motion blur and temporal inconsistencies, suggesting poor editing quality. Preservation is low because unrelated elements (e.g., background people, signs) are distorted or blurred, and realism is poor due to unnatural motion artifacts. Consistency is also low due to temporal drift and inconsistent frame rates."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 1,\n \"consistency\": 1,\n \"reason\": \"The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks are actually faster and more blurred, indicating motion acceleration rather than slowing down. This is a critical accuracy failure. Additionally, background elements like the fan and signage show motion blur and temporal inconsistencies, suggesting poor editing quality. Preservation is low because unrelated elements (e.g., background people, signs) are distorted or blurred, and realism is poor due to unnatural motion artifacts. Consistency is also low due to temporal drift and inconsistent frame rates.\"\n}",
+ "receipt_digest": "40124c033544851b509f9434b495b8d8b089cb26d5f70898a2fad54147173683",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 1.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 1.0,
+ "reason": "The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks are actually faster and more blurred, indicating motion acceleration rather than slowing down. This is a critical accuracy failure. Additionally, background elements like the fan and signage show motion blur and temporal inconsistencies, suggesting poor editing quality. Preservation is low because unrelated elements (e.g., background people, signs) are distorted or blurred, and realism is poor due to unnatural motion artifacts. Consistency is also low due to temporal drift and inconsistent frame rates."
+ },
+ "source_video_sha256": "89033eb8d081cf564f28ff41111f25fcd6bd7f41604c0f4d3e54bb6f950f5ee8",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/liveedit_wan21_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/liveedit_wan21_1p3b.json
new file mode 100644
index 0000000..2fa1fd4
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/liveedit_wan21_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T20:10:55Z",
+ "edited_video_sha256": "920905582914e55b249fb889a697a64f987e527e4583fb70f912d6ce4d664164",
+ "editing_instruction": "Slow down the girl's kicking motion.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "fc731a49e6c3d35b5945782f04ebc463488a0e3e5ede54e972425835c208e5d7",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 85,
+ "item_id": "Te_Motion_frequency_change_06",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 2,
+ "preservation": 2,
+ "realism": 3,
+ "reason": "The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks are actually faster and more exaggerated, with unnatural motion blur and orange-colored feet, indicating a complete failure in temporal attribute correction. Additionally, the girl's belt color changed from black to brown, and other students' uniforms appear altered, violating preservation. The motion is unrealistic and inconsistent across frames, with visible artifacts and temporal drift."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 3,\n \"consistency\": 2,\n \"reason\": \"The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks are actually faster and more exaggerated, with unnatural motion blur and orange-colored feet, indicating a complete failure in temporal attribute correction. Additionally, the girl's belt color changed from black to brown, and other students' uniforms appear altered, violating preservation. The motion is unrealistic and inconsistent across frames, with visible artifacts and temporal drift.\"\n}",
+ "receipt_digest": "763853f76cab0cd215dcb247836866d9626f7135ddf034171a78ef0d112836ed",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 2.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 3.0,
+ "reason": "The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks are actually faster and more exaggerated, with unnatural motion blur and orange-colored feet, indicating a complete failure in temporal attribute correction. Additionally, the girl's belt color changed from black to brown, and other students' uniforms appear altered, violating preservation. The motion is unrealistic and inconsistent across frames, with visible artifacts and temporal drift."
+ },
+ "source_video_sha256": "89033eb8d081cf564f28ff41111f25fcd6bd7f41604c0f4d3e54bb6f950f5ee8",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/lucy_edit_1p1_dev_5b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/lucy_edit_1p1_dev_5b.json
new file mode 100644
index 0000000..d94af20
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/lucy_edit_1p1_dev_5b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T18:19:12Z",
+ "edited_video_sha256": "8cdbcf5381196d017811d9f74464977bc946bb670dd9fb07d60b46b95278ee31",
+ "editing_instruction": "Slow down the girl's kicking motion.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "958406a72abc3d1d3ccff777bbe657b4e77061cb2ecce850fc4a0f8546716c3a",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 85,
+ "item_id": "Te_Motion_frequency_change_06",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 4,
+ "preservation": 5,
+ "realism": 3,
+ "reason": "The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks at the same or even faster speed than the original. The motion is not slowed down, making the edit incorrect. Preservation is intact as background elements and other people remain unchanged. Realism is acceptable with no major artifacts, and consistency is good with no temporal drift. However, since the core instruction is not followed, accuracy is 0."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 5,\n \"realism\": 3,\n \"consistency\": 4,\n \"reason\": \"The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks at the same or even faster speed than the original. The motion is not slowed down, making the edit incorrect. Preservation is intact as background elements and other people remain unchanged. Realism is acceptable with no major artifacts, and consistency is good with no temporal drift. However, since the core instruction is not followed, accuracy is 0.\"\n}",
+ "receipt_digest": "7cf0fd36b9952a98b01afed58f37c829551aee427c5cff16d0284fceb365cafe",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 4.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 5.0,
+ "realism": 3.0,
+ "reason": "The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks at the same or even faster speed than the original. The motion is not slowed down, making the edit incorrect. Preservation is intact as background elements and other people remain unchanged. Realism is acceptable with no major artifacts, and consistency is good with no temporal drift. However, since the core instruction is not followed, accuracy is 0."
+ },
+ "source_video_sha256": "89033eb8d081cf564f28ff41111f25fcd6bd7f41604c0f4d3e54bb6f950f5ee8",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/omnivideo2_official_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/omnivideo2_official_1p3b.json
new file mode 100644
index 0000000..ce45e18
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/omnivideo2_official_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T18:23:29Z",
+ "edited_video_sha256": "0ec0232e51cbdbced06ee86e9c81712facbe845ba0689521e2f45547925dd786",
+ "editing_instruction": "Slow down the girl's kicking motion.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "1b79bbf410c6adf53abd87332bc0cc6930b4f54494ede953234b3355e38afefb",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 85,
+ "item_id": "Te_Motion_frequency_change_06",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 1,
+ "preservation": 2,
+ "realism": 1,
+ "reason": "The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks are not slowed down — in fact, they appear distorted, fragmented, and unnatural with visible motion artifacts and ghosting. The motion is not temporally coherent, and the girl's body parts blur and warp in ways that suggest a failed or incorrect edit. Background elements (other students, banners) are also distorted, indicating the edit affected unrelated regions. This is a major failure in accuracy and realism, with poor preservation and consistency."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 1,\n \"consistency\": 1,\n \"reason\": \"The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks are not slowed down — in fact, they appear distorted, fragmented, and unnatural with visible motion artifacts and ghosting. The motion is not temporally coherent, and the girl's body parts blur and warp in ways that suggest a failed or incorrect edit. Background elements (other students, banners) are also distorted, indicating the edit affected unrelated regions. This is a major failure in accuracy and realism, with poor preservation and consistency.\"\n}",
+ "receipt_digest": "c1de92d68a3494fe138696fe42ff1694127fcdee348e4320b549f0c5d59a6aa0",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 1.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 1.0,
+ "reason": "The instruction was to slow down the girl's kicking motion, but the edited video shows her kicks are not slowed down — in fact, they appear distorted, fragmented, and unnatural with visible motion artifacts and ghosting. The motion is not temporally coherent, and the girl's body parts blur and warp in ways that suggest a failed or incorrect edit. Background elements (other students, banners) are also distorted, indicating the edit affected unrelated regions. This is a major failure in accuracy and realism, with poor preservation and consistency."
+ },
+ "source_video_sha256": "89033eb8d081cf564f28ff41111f25fcd6bd7f41604c0f4d3e54bb6f950f5ee8",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/reco_ori_wan21_1p3b.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/reco_ori_wan21_1p3b.json
new file mode 100644
index 0000000..df18015
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/reco_ori_wan21_1p3b.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T20:08:12Z",
+ "edited_video_sha256": "e59014ce6f558d6fb175cac92be6a90e8af459a6bc05eb255878ce42821a007b",
+ "editing_instruction": "Slow down the girl's kicking motion.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "ce02857d7678e154c3e4ed0e6feab8073b9587fe0c17550586d549a5e1bf13e5",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 85,
+ "item_id": "Te_Motion_frequency_change_06",
+ "parsed_judge_output": {
+ "accuracy": 0,
+ "consistency": 1,
+ "preservation": 2,
+ "realism": 1,
+ "reason": "The instruction was to slow down the girl's kicking motion, but the edited video completely removes the girl from the scene and replaces her with a different child who is not kicking at all. This is a fundamental failure in accuracy. The background elements are mostly preserved, but there are noticeable inconsistencies (e.g., the child in the background changes position and behavior unnaturally). The motion of the remaining child is not slowed down, and the overall realism is poor due to the abrupt removal of the main subject and unnatural behavior of the replacement child. Temporal consistency is broken as the action sequence is disrupted."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 0,\n \"preservation\": 2,\n \"realism\": 1,\n \"consistency\": 1,\n \"reason\": \"The instruction was to slow down the girl's kicking motion, but the edited video completely removes the girl from the scene and replaces her with a different child who is not kicking at all. This is a fundamental failure in accuracy. The background elements are mostly preserved, but there are noticeable inconsistencies (e.g., the child in the background changes position and behavior unnaturally). The motion of the remaining child is not slowed down, and the overall realism is poor due to the abrupt removal of the main subject and unnatural behavior of the replacement child. Temporal consistency is broken as the action sequence is disrupted.\"\n}",
+ "receipt_digest": "20806937247385e596fc1dfbc867e577e3c23a7a421ce5113999832756cfbf18",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 0.0,
+ "accuracy_penalty": 0.0,
+ "consistency": 1.0,
+ "overall_score_paper_0_to_100": 0.0,
+ "overall_score_raw_0_to_5": 0.0,
+ "preservation": 2.0,
+ "realism": 1.0,
+ "reason": "The instruction was to slow down the girl's kicking motion, but the edited video completely removes the girl from the scene and replaces her with a different child who is not kicking at all. This is a fundamental failure in accuracy. The background elements are mostly preserved, but there are noticeable inconsistencies (e.g., the child in the background changes position and behavior unnaturally). The motion of the remaining child is not slowed down, and the overall realism is poor due to the abrupt removal of the main subject and unnatural behavior of the replacement child. Temporal consistency is broken as the action sequence is disrupted."
+ },
+ "source_video_sha256": "89033eb8d081cf564f28ff41111f25fcd6bd7f41604c0f4d3e54bb6f950f5ee8",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/vace_wan21_1p3b_fullframe.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/vace_wan21_1p3b_fullframe.json
new file mode 100644
index 0000000..ae5e653
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/vace_wan21_1p3b_fullframe.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-14T19:58:58Z",
+ "edited_video_sha256": "d54f0a371d7519fd205d5fb20f38520953477a5d6bd61d807e58d735861f4c4f",
+ "editing_instruction": "Slow down the girl's kicking motion.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "dd144add239638ebbc6a7f5dd0112fdc48e4b0eb193048b94cd97eecdeacb179",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 85,
+ "item_id": "Te_Motion_frequency_change_06",
+ "parsed_judge_output": {
+ "accuracy": 2,
+ "consistency": 2,
+ "preservation": 3,
+ "realism": 2,
+ "reason": "The girl's kicking motion is not meaningfully slowed down; the speed appears nearly identical to the original, failing the core instruction. The motion is slightly distorted with unnatural artifacts (e.g., leg blur, inconsistent foot position), indicating poor temporal interpolation. Background elements (other students, signs) are mostly preserved but show minor color shifts and slight warping. Temporal consistency is poor—motion appears jerky and inconsistent across frames, with visible drift in the girl’s body position. Overall, the edit does not achieve the requested slow-down and introduces visual artifacts."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 2,\n \"preservation\": 3,\n \"realism\": 2,\n \"consistency\": 2,\n \"reason\": \"The girl's kicking motion is not meaningfully slowed down; the speed appears nearly identical to the original, failing the core instruction. The motion is slightly distorted with unnatural artifacts (e.g., leg blur, inconsistent foot position), indicating poor temporal interpolation. Background elements (other students, signs) are mostly preserved but show minor color shifts and slight warping. Temporal consistency is poor—motion appears jerky and inconsistent across frames, with visible drift in the girl’s body position. Overall, the edit does not achieve the requested slow-down and introduces visual artifacts.\"\n}",
+ "receipt_digest": "b8e9b25e57cfd4c0945db065483ea53642a5be449c30f31e7445a030e58a3755",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 2.0,
+ "accuracy_penalty": 0.4,
+ "consistency": 2.0,
+ "overall_score_paper_0_to_100": 29.6,
+ "overall_score_raw_0_to_5": 1.48,
+ "preservation": 3.0,
+ "realism": 2.0,
+ "reason": "The girl's kicking motion is not meaningfully slowed down; the speed appears nearly identical to the original, failing the core instruction. The motion is slightly distorted with unnatural artifacts (e.g., leg blur, inconsistent foot position), indicating poor temporal interpolation. Background elements (other students, signs) are mostly preserved but show minor color shifts and slight warping. Temporal consistency is poor—motion appears jerky and inconsistent across frames, with visible drift in the girl’s body position. Overall, the edit does not achieve the requested slow-down and introduces visual artifacts."
+ },
+ "source_video_sha256": "89033eb8d081cf564f28ff41111f25fcd6bd7f41604c0f4d3e54bb6f950f5ee8",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/zebra.json b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/zebra.json
new file mode 100644
index 0000000..b7e97ae
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/evidence/Te_Motion_frequency_change_06/zebra.json
@@ -0,0 +1,46 @@
+{
+ "category": "Motion Attribute",
+ "created_at_utc": "2026-09-22T10:01:55Z",
+ "edited_video_sha256": "99e7098bf8ca8f327d9acf1c72b3c88b82ca7e79c3b7fc22c0fd503f53a88f23",
+ "editing_instruction": "Slow down the girl's kicking motion.",
+ "evaluator": {
+ "do_sample": false,
+ "frame_count_per_video": 12,
+ "frame_sampling": "uniform_interval_midpoints",
+ "input_manifest_sha256": "149da52dc20cd8ac0dcb568d1eb13e785bc243ac1c0d1be8003e09e0e9938f7b",
+ "judge": "Qwen3-VL-32B-Instruct",
+ "model_config_sha256": "d2dd0c60d01b9e195d9447c52da61c7302d28828524914c044d9c6e1b81d0427",
+ "model_index_sha256": "06c952c569285870b811989b794b9766493e280fb77fbcb957fc4e5fcf25403a",
+ "model_path": "/vast/users/guangyi.chen/causal_group/peiyuan.zhu/work/VLM/MEV-Annotation/checkpoints/Qwen3-VL-32B-Instruct",
+ "official_omni_score": false,
+ "prompt_sha256": "8168fe4b4edf0f608e79476fd6b9edf28cb1ede6e6734184f1b12d65c2dedbaf",
+ "temperature": 0.0,
+ "transformers_version": "5.5.4",
+ "wrapper_sha256": "c087ecfdf0fedf00e224ad1795ee87a4f7e01fb4875bcc35537dd1240e5d6dd4"
+ },
+ "index": 85,
+ "item_id": "Te_Motion_frequency_change_06",
+ "parsed_judge_output": {
+ "accuracy": 5,
+ "consistency": 5,
+ "preservation": 5,
+ "realism": 5,
+ "reason": "The girl's kicking motion is accurately slowed down while maintaining natural motion and timing. All other elements in the video, including background people, camera position, and environment, remain unchanged and consistent. The motion is smooth with no artifacts, and temporal stability is preserved throughout the sequence. The edit strictly follows the instruction without introducing any errors."
+ },
+ "protocol_id": "omni_temporal200",
+ "raw_judge_output": "{\n \"accuracy\": 5,\n \"preservation\": 5,\n \"realism\": 5,\n \"consistency\": 5,\n \"reason\": \"The girl's kicking motion is accurately slowed down while maintaining natural motion and timing. All other elements in the video, including background people, camera position, and environment, remain unchanged and consistent. The motion is smooth with no artifacts, and temporal stability is preserved throughout the sequence. The edit strictly follows the instruction without introducing any errors.\"\n}",
+ "receipt_digest": "0f6bc3a22fd9d02baf2933af9d37ab636676f15d25234251b1a65d1a7d45c38d",
+ "schema_version": "action-editing-omni-prompt-qwen3vl-result-v1",
+ "scores": {
+ "accuracy": 5.0,
+ "accuracy_penalty": 1.0,
+ "consistency": 5.0,
+ "overall_score_paper_0_to_100": 100.0,
+ "overall_score_raw_0_to_5": 5.0,
+ "preservation": 5.0,
+ "realism": 5.0,
+ "reason": "The girl's kicking motion is accurately slowed down while maintaining natural motion and timing. All other elements in the video, including background people, camera position, and environment, remain unchanged and consistent. The motion is smooth with no artifacts, and temporal stability is preserved throughout the sequence. The edit strictly follows the instruction without introducing any errors."
+ },
+ "source_video_sha256": "89033eb8d081cf564f28ff41111f25fcd6bd7f41604c0f4d3e54bb6f950f5ee8",
+ "subcategory": "Motion_frequency_change"
+}
diff --git a/projects/video-action-editing/qualitative/omni/index.html b/projects/video-action-editing/qualitative/omni/index.html
new file mode 100644
index 0000000..e3df2d1
--- /dev/null
+++ b/projects/video-action-editing/qualitative/omni/index.html
@@ -0,0 +1,93 @@
+
+
+
+
+Omni Temporal · Zebra Top 5
+
+
+
+
+
+
Qualitative comparison / September 2026BEST-CASE SHOWCASE
is independent of the initial state, action regime 
, and innovations. The renderer combines content with the full realized trajectory 
.
is injective on a full-measure action subset; 
is the action-conditional kernel mean embedding. Individual semantic coordinates are not fixed.
with symmetric nonzero content is invertible, yet ![E[V | z^a] = 0](./paper/equations/zero-conditional-mean.svg)
. Predicting this mean can erase all action differences.
Normalized anchor
![Δu = α_k W_up [(W_u u) ⊙ tanh(W_k k) + (W_k k) ⊙ (1 + tanh(W_b b))].](./paper/equations/adapter-update.svg)
also includes V-JEPA2 features.
![ℓ_act,k = 1 − γ_{+,k} − λ_disc log [exp(γ_{+,k}/τ) / Σ_{j∈K} exp(γ_{j,k}/τ)].](./paper/equations/behavior-loss.svg)
