Pilot-80 · H3 Omni Ref

Final runner-ready set · updated 2026-08-31 02:39 PDT · manual refresh only

80 / 80unique source combinations
80 / 80H3 outputs synchronized
51 / 80raw H3 accepted
1raw outputs needing review

Distribution: A1 28 · A3 4 · A4 24 · D4 24. A3 is a capability canary and is not allowed to inflate the core promotion denominator.

Cluster: not queued

Raw H3 QA: 80 audited, 51 accepted, 1 review, 28 rejected, 19 with unexpected or unintelligible speech.

Delivery: original silent contracts are stripped deterministically. Generated ambience is retained unless raw QA confirms unexpected or unintelligible speech; those outputs receive a separately labeled speech-sanitized delivery. Delivery and visual salvage never count as raw model success.

Evidence-22 · All rounds

#01 · ref2va-sample-2026083182-000000

A1 · 10s

eligiblecompletedraw rejecteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.85

The visual generation matches the input reference and prompt requirements well, but unexpected unintelligible speech violates the no_speech directive.

task 3/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 2/4 · artifact 4/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately depicts the scene from the reference image with smooth motion and faithful subject and environment retention.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 10-second video of the man from the picture standing beside the large aquarium, calmly watching the colorful discus fish swim around and gently tapping on the glass without speaking.

Exact H3 prompt
In a realistic indoor setting, generate a 10-second continuous shot of the man from <Picture 1> wearing his black baseball cap and white shirt standing beside the large illuminated aquarium. He quietly watches the vibrant red, orange, and patterned discus fish glide through the water, gently tapping his fingers against the glass as the fish swim closer. His mouth remains closed and naturally at rest throughout. Ambient room tone with subtle aquarium filter hum and water bubbling is heard.
#02 · ref2va-sample-2026083182-000001

A1 · 12s

eligiblecompletedraw reviewdelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

REVIEW · confidence 0.82

The generated video accurately reproduces the habitat from the reference image and animates the hamster exploring and mounting the wheel in compliance with the prompt instructions.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated video successfully recreates the container habitat from the reference and depicts the requested hamster motion naturally while following audio constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a realistic 12-second video of a cute little hamster exploring the clear plastic bin cage setup from the photo, digging in the wood shavings and hopping onto the wheel to run.

Exact H3 prompt
A realistic 12-second video shot from a medium high angle of a small golden Syrian hamster exploring the clear plastic bin habitat shown in <Picture 1>. The hamster emerges from the wood shavings bedding, sniffs around with twitching whiskers, climbs into the clear exercise wheel from <Picture 1>, and runs smoothly as the wheel spins. The hamster keeps its mouth closed with no human speech. Audio consists of the soft rustling of wood shavings, the gentle spinning whir of the wheel, and quiet room ambience.
#03 · ref2va-sample-2026083182-000002

A1 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video effectively demonstrates peeling away scrap dough from around the fish-shaped pastry cutout shown in the reference image, adhering to the requested actions and constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video successfully executes the requested action of peeling dough scraps away from the fish cutout with good visual fidelity and coherent motion.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Show hands carefully peeling away the excess dough around the fish cutout from the floured counter.

Exact H3 prompt
A top-down close-up shot of hands gently peeling away and removing the scrap dough surrounding the fish-shaped pastry dough shown in <Picture 1>. The clean fish dough cutout remains resting undisturbed on the dark, lightly floured work surface. Overhead lighting with subtle dough-handling Foley, no dialogue or speech.
#04 · ref2va-sample-2026083182-000003

A1 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video accurately translates the reference image subjects into a continuous walking motion along the garden path with consistent identity, coherent motion, and matching ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated video accurately recreates the subject, dog, and garden setting from the reference image, performing the requested walking action smoothly with appropriate ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Generate a video of the woman from the picture walking her small dog down the garden path with her basket.

Exact H3 prompt
A medium-wide shot showing the woman and small puppy from <Picture 1> walking forward along the dirt garden path surrounded by leafy vegetable patches. The woman carries a basket in hand and holds the puppy's leash as the puppy trots alongside her. Her mouth remains closed and at rest with no spoken dialogue. The camera tracks backward at a gentle, steady pace to keep them in frame under bright natural daylight. Natural outdoor ambience with rustling leaves, distant chirping birds, and soft footsteps on dirt.
#05 · ref2va-sample-2026083182-000004

A1 · 10s

eligiblecompletedraw rejecteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.85

The generated video accurately replicates the subject's identity and attire while faithfully executing the requested runway motion, pose, and turn without speech.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video successfully portrays the subject from Picture 1 performing the full runway sequence with high visual fidelity and compliant audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 10-second fashion runway video featuring the woman from Picture 1. Have her walk forward down the runway toward the camera with a confident model strut, pause smoothly at the front mark, pose with one hand on her hip, and turn gracefully to walk back up the runway. Keep the atmosphere professional with ambient runway lighting and camera flash effects, with no spoken dialogue.

Exact H3 prompt
Generate a 10-second video of the woman from <Picture 1> walking down a high-fashion runway. She strides forward confidently toward the camera in her matching turquoise outfit and pink platform heels, her mouth closed with a composed runway expression. At 00:04, she reaches the end of the catwalk, pauses, rests one hand gracefully against her hip, and looks directly forward while subtle camera flashes illuminate the background. At 00:07, she pivots smoothly on her heels and walks away up the catwalk with rhythmic strides. The camera tracks backward slightly as she approaches, holding a medium-full shot. No spoken dialogue; ambient runway room tone and rhythmic high-heel footsteps on the glossy catwalk.
#06 · ref2va-sample-2026083182-000005

A1 · 10s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video successfully preserves the subject identity and outfit while executing the requested motion and studio setting.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.86

The model successfully generated the requested video adhering to the subject reference, environment, motion prompt, and audio constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 10-second video of the woman from Picture 1 in a bright, modern studio. She moves gracefully, shifts her posture, and looks toward the camera with a calm expression. Her mouth stays closed with no dialogue, accompanied only by quiet studio room tone.

Exact H3 prompt
In a clean, brightly lit modern studio with soft daylight, the woman from <Picture 1> stands centered in her off-white lace-trimmed pleated dress and black open-toe booties. She takes a slow step forward, pauses, and gently turns her shoulders while smoothing the side of her dress. She looks directly toward the camera with a serene, poised expression. The camera performs a slow eye-level push-in. Her mouth remains closed and at rest throughout with no lip movement, dialogue, or vocalization. Quiet indoor studio ambience and subtle fabric rustling are heard.
#07 · ref2va-sample-2026083182-000006

A1 · 12s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video accurately preserves the model identity and clothing while executing natural studio modeling motions with matching audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately captures the model identity, attire, and requested studio posing motions with high visual and temporal consistency.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 12-second fashion studio video featuring the model from the photo. Have her strike dynamic modeling poses, shifting her stance, adjusting her blazer, and looking confidently into the camera against a bright cyclorama studio background. No talking or dialogue, just studio sound and fashion background music.

Exact H3 prompt
A 12-second fashion video featuring the woman from <Picture 1> in a seamless white studio cyclorama. She smoothly alternates modeling poses, shifting weight, adjusting her black blazer, and gazing confidently into the lens. Her mouth remains closed and naturally at rest throughout, with no lip movement, no dialogue, no vocalization, and no human voice. Camera performs smooth dolly push-ins and framing adjustments. Ambient sound includes studio room tone, soft fabric rustle, and light heel clicks over an upbeat electronic fashion music track.
#08 · ref2va-sample-2026083182-000007

A1 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video accurately animates the subject from Picture 1 performing the runway walk and pose with consistent identity and appropriate non-speech audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The video accurately animates the model from Picture 1 performing a runway walk and pose with consistent identity, smooth motion, and compliant ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 5-second video of the fashion model from Picture 1 walking confidently down the runway, striking a poised pose, and beginning a smooth pivot, with ambient venue sound and no spoken dialogue.

Exact H3 prompt
Generate a 5-second video of the woman from <Picture 1> walking forward down a brightly lit fashion show runway with a poised, confident runway stride. The camera slowly tracks backward at full-body framing as she walks forward, pauses briefly at the mark to strike an elegant pose, and smoothly pivots. Ambient fashion show venue sound and subtle shoe clicks on the runway floor; no spoken dialogue and her mouth remains closed.
#09 · ref2va-sample-2026083182-000008

A1 · 5s

eligiblecompletedraw rejecteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.85

Visual identity preservation and motion transfer are executed well, but the generation fails due to unexpected speech violating the no_speech policy.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 2/4 · artifact 4/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.9

The generated video accurately depicts the subject performing the requested actions in the garage setting with natural motion and strong identity preservation.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 5-second realistic video of the mechanic from the photo standing in an auto workshop, calmly looking around the garage and crossing his arms across his chest.

Exact H3 prompt
Generate a 5-second realistic video featuring the man from <Picture 1> standing inside an auto repair garage. His mouth remains closed and at rest throughout with no spoken dialogue. He looks calmly around the workshop bay before shifting his arms forward to cross them over his chest in a steady, confident posture, captured in a subtle slow push-in shot with natural ambient garage sound.
#10 · ref2va-sample-2026083182-000009

A1 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The generated video accurately animates the subject from Picture 1 with high visual fidelity, natural movement, and adherence to the no-speech condition.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.86

The generated video successfully animates the subject from Picture 1 following all prompt instructions with natural motion and correct audio policy compliance.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 5-second realistic video of the young man in the white shirt and glasses from Picture 1 standing on the entrance steps, holding the boxed bouquet with both hands, smiling, and giving a gentle friendly nod without speaking.

Exact H3 prompt
A 5-second realistic video of the young man from <Picture 1>, wearing black-framed glasses, a white long-sleeve button-down shirt, and dark trousers. Standing on polished marble entrance steps, he holds the clear box containing a floral bouquet in both hands. He smiles warmly, shifts his weight slightly, and gives a polite, welcoming nod. The camera performs a slow, subtle push-in in a medium shot. His mouth remains closed and naturally at rest with no lip movement, no dialogue, no vocalization, and no human voice, accompanied only by faint outdoor ambient sound.
#11 · ref2va-sample-2026083182-000010

A1 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The generated video accurately translates the reference image into the requested runway motion sequence with consistent subject appearance and appropriate ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated video successfully animates the subject performing a realistic runway turn while faithfully preserving appearance, clothing, and audio constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Can you generate a 5-second video of the woman from the photo posing and turning smoothly on a fashion runway under bright stage lights without speaking?

Exact H3 prompt
A 5-second realistic fashion runway video featuring the woman from <Picture 1>. She stands on the black runway under bright spotlights, places her hand on her hip, and executes a smooth, elegant pivot turn to showcase her outfit before settling into a confident forward pose. Her mouth remains closed and at rest throughout with no spoken dialogue or vocalization. Ambient runway hall acoustics and faint floor scuffs are heard.
#12 · ref2va-sample-2026083182-000011

A1 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.85

The generated video accurately animates the subject from Picture 1 with high identity fidelity, natural motion, and compliant audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated video accurately brings the reference subject to life with natural studio posing, consistent attire, and appropriate audio policy compliance.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Make a short 5-second video of the girl from the photo posing and smiling at the camera in a photo studio.

Exact H3 prompt
A 5-second studio video of the woman from <Picture 1>. In a static full-body shot against a seamless white studio backdrop, she wears her olive-green tank top, distressed light-blue jeans, and brown lace-up shoes. She shifts her weight gently, rests one hand near her hip pocket, lightly brushes her hair back with the other hand, and smiles warmly at the camera. Her mouth remains closed and at rest throughout, with no dialogue or speech. Faint studio room tone and soft fabric movement are audible.
#13 · ref2va-sample-2026083182-000012

A1 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.9

The generated video accurately portrays the reference subject posing at a press event with smooth motion and appropriate ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.85

The generated video accurately follows the prompt instructions, maintaining character fidelity, realistic subtle motion, and appropriate sound design.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Make a 5-second video of the woman in the picture posing for photos at a press event, smiling and shifting her pose slightly under flashing cameras.

Exact H3 prompt
Generate a 5-second video featuring the woman from <Picture 1> standing in front of an event backdrop. She maintains her exact appearance, hairstyle, and patterned purple chiffon dress from <Picture 1>. She poses elegantly for the press, resting one hand on her hip, adjusting her posture slightly, and turning her head subtly to smile at different photographers. Camera flashes intermittently illuminate the scene. Her mouth remains closed in a gentle smile throughout with no spoken dialogue. The shot is a steady full-length framing with subtle slow push-in. Audio consists of ambient room tone, soft murmuring, and the rhythmic clicking of camera shutters.
#14 · ref2va-sample-2026083182-000013

A1 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.85

The video faithfully reproduces the character and attire from the reference image, successfully executing the notebook interaction and camera interaction smoothly.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated video accurately depicts the subject from Picture 1 performing the requested actions with high visual and temporal fidelity.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Make an 8-second video of the 18th-century man from the picture standing outside an old manor, opening his notebook, flipping through a couple of pages, and smiling thoughtfully at the camera.

Exact H3 prompt
In an 8-second video, the man in 18th-century attire from <Picture 1> stands on a historic stone terrace in front of a classical building. Holding the small notebook from <Picture 1>, he gently opens it, leafs through a page with his thumb, and looks up with a pleasant, thoughtful smile. The camera slowly pushes in from a medium-full shot to a medium shot in soft natural daylight. His mouth remains closed with no spoken dialogue. Gentle ambient outdoor courtyard atmosphere with faint paper rustle and fabric movement.
#15 · ref2va-sample-2026083182-000014

A1 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video successfully animates the subject from Picture 1 with consistent identity and natural red-carpet posing motions while adhering to audio and duration requirements.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated video accurately depicts the subject from Picture 1 posing on the red carpet with natural motion, matching appearance, and clean ambient audio adhering to the no_speech requirement.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Make an 8-second video of the young man in the navy suit from the photo posing and smiling on the red carpet for photographers without speaking.

Exact H3 prompt
An 8-second video of the young man from <Picture 1> posing on a red carpet at a media event. He stands in his navy suit and white undershirt, subtly shifting his weight from one foot to the other and turning his head from side to side to acknowledge off-screen photographers with a calm, confident smile. His mouth remains closed and at rest throughout with no speaking. The camera maintains a medium-full framing with a subtle slow push-in. Generated ambient soundscape features distant crowd murmur and camera shutter clicks.
#16 · ref2va-sample-2026083182-000015

A1 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The generated video accurately incorporates the reference subject's identity and gear into the requested dock environment with natural movement and fitting ambient ocean audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.86

The generated video accurately incorporates the reference subject into the requested coastal dock setting with appropriate motion and clean ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Make an 8-second video of the diver from <Picture 1> standing on a sunny wooden dock by the sea, checking his snorkeling gear and looking around.

Exact H3 prompt
Generate an 8-second realistic video featuring the man in snorkeling gear from <Picture 1>. He stands on a sunlit wooden dock overlooking calm turquoise ocean water. He adjusts the strap of his snorkeling mask hanging from his vest, looks out across the water with a calm smile, and rests his hands at his side. The camera maintains a medium full-body framing with a gentle slow push-in. Ambient sounds of gentle sea waves lapping against the dock and distant coastal breeze. No speech or vocalization.
#17 · ref2va-sample-2026083182-000016

A1 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The model accurately animated the subject from Picture 1 performing a runway walk and pose with consistent identity, coherent motion, and compliant audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video closely adheres to the reference subject and instructions, featuring consistent motion, realistic staging, and compliant audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create an 8-second realistic video of the woman from the photo walking confidently down an outdoor fashion runway surrounded by green trees, pausing at the front of the stage to strike a poised pose, and smoothly turning around without speaking.

Exact H3 prompt
An 8-second realistic outdoor fashion video featuring the woman from <Picture 1>. She walks smoothly along a dark outdoor runway stage surrounded by leafy green park trees under soft daylight. She reaches the front of the runway, halts, and rests one hand gracefully on her hip while striking a poised high-fashion pose before executing a clean pivot turn. Her mouth remains closed throughout with a composed expression and no dialogue. Filmed with a smooth tracking camera dollied back ahead of her walk that gently settles during her pose. Ambient outdoor breeze, soft rustling foliage, and quiet footstep Foley on the stage.
#18 · ref2va-sample-2026083182-000017

A1 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.9

The video accurately animates the subject from Picture 1 in the specified beach environment with natural subtle motion and fitting ambient sound.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately depicts the subject from Picture 1 animated naturally on the beach with matching ambient audio and no speech.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create an 8-second realistic video of the woman in the white dress from the picture standing on the beach, holding her bouquet of orange flowers and smiling gently as the ocean breeze moves her hair.

Exact H3 prompt
In an 8-second shot on a sunny beach, the woman from <Picture 1> stands facing the camera holding her orange floral bouquet. Coastal breeze gently moves her hair and white dress while turquoise waves break behind her. She smiles softly with her mouth closed and naturally at rest, with no lip movement, no dialogue, no vocalization, and no human voice. The camera holds a steady medium full-body framing. Ambient ocean waves and gentle wind.
#19 · ref2va-sample-2026083182-000018

A1 · 10s

eligiblecompletedraw rejecteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.95

The visual animation accurately renders the character from Picture 1 eating and reacting, but the output fails due to unexpected unintelligible vocalizations violating the no-speech constraint.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 2/4 · artifact 4/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately captures the character identity and requested sequence of actions with consistent animation and good visual quality.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create an anime scene of the orange-haired boy from Picture 1 enjoying his spicy lunch outdoors by the flower garden, reacting playfully to the spice and taking a sip of water.

Exact H3 prompt
In vibrant anime style, the orange-haired boy from <Picture 1> sits outdoors by blooming flower bushes. He eats a spoonful of spicy tofu from his dish, chews with closed lips, reacts playfully to the spice with a flushed grin, and sips from a glass of iced water. No speech, no dialogue, no vocalization, and no background music; only gentle garden ambience with subtle spoon and glass clinks.
#20 · ref2va-sample-2026083182-000019

A1 · 10s

eligiblecompletedraw rejecteddelivery rejected

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.85

The generation captures the character identity and dance well, but contains audible singing/speech violating the no-speech policy and shows occasional visual artifacts.

task 3/4 · preserve 3/4 · identity 4/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: unexpected_speech, visible_artifacts

Delivery output

speech removed after raw failure

Delivery QA

REJECTED · confidence 0.85

The generated video accurately depicts the character performing on stage with faithful identity preservation and smooth idol choreography.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 3/4

failure codes: none

visual salvage: false

Natural request

Generate a 10-second video of the anime idol character from <Picture 1> performing an energetic dance routine on a brightly lit concert stage, executing cute choreography and waving to the crowd with no spoken dialogue.

Exact H3 prompt
A vibrant anime-style video of the anime idol character from <Picture 1> performing on a bright concert stage decorated with colorful balloons and glowing light panels. The character from <Picture 1>, dressed in the identical light blue and white sailor stage costume with cap and red neckerchief, moves energetically across the stage, executing lively idol dance steps, waving toward the audience, and finishing with an enthusiastic pose. The camera tracks smoothly in medium shot with dynamic zooms. No spoken dialogue; mouth remains in a cheerful resting expression with upbeat stage ambience and cheering crowd sounds.
#21 · ref2va-sample-2026083182-000020

A1 · 12s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video closely follows the prompt instructions, maintaining subject identity, requested camera angles, and atmospheric audio without spoken dialogue.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated video closely adheres to the reference image and prompt instructions, demonstrating consistent identity, clean animation, and accurate timing.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 12-second cinematic anime video featuring the character from <Picture 1> standing on a rooftop at night overlooking the city. Have the night wind blow through her long hair and coat as she turns and looks thoughtfully over her shoulder, keeping her expression calm and silent.

Exact H3 prompt
Generate a 12-second cinematic anime scene featuring the anime girl from <Picture 1> on a rooftop overlooking a glowing cityscape at night. The camera starts in a medium shot showing her standing near a chain-link fence as a steady breeze ripples through her long black hair and the hem of her dark coat. She gazes back over her shoulder with an introspective expression, keeping her mouth closed and resting. At 00:06.000, cut to a close-up tracking her face as she slowly shifts her amber eyes toward the distant city lights, her hair strands drifting across the frame. Ambient rooftop wind and distant low city hum play throughout with no spoken dialogue.
#22 · ref2va-sample-2026083182-000021

A1 · 12s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video successfully brings the reference character to life with faithful visual styling, smooth choreography matching all prompt instructions, and appropriate ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated animation accurately reproduces the character identity and environment from the reference image, successfully executing the requested bow and gesture sequence with clean motion and compliant audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 12-second anime animation of the elegant dark-haired nobleman from <Picture 1> in his lavish ballroom setting. Have him perform a smooth, theatrical welcoming bow and elegant hand gestures with a composed, confident smirk. Keep his mouth closed throughout with no spoken dialogue.

Exact H3 prompt
Generate a 12-second anime sequence featuring the dark-haired nobleman from <Picture 1> in his grand crimson and gold parlor. He begins in a welcoming pose with an extended gloved hand, then executes a slow, theatrical bow while resting his left hand over his heart. He rises gracefully, smoothly adjusting his glove and offering a confident, enigmatic smirk toward the camera. Soft golden sparkles drift through the ambient light. His mouth remains closed throughout with no speech or lip motion. Gentle cloth rustling and a rich orchestral chamber ambient room tone accompany the scene.
#23 · ref2va-sample-2026083182-000022

A1 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The model accurately animates the sketch into a clean landing sequence adhering to the identity reference, timing, and audio requirements without speech.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately animates the stylized sketch character performing a dynamic landing and ready stance while maintaining style consistency and audio constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Animate the stylized character from the sketch in <Picture 1> performing a dynamic anime battle landing against a clean white backdrop. She drops down, lands lightly on her feet, and gets into a ready stance while her creature-like hair moves independently. No spoken dialogue.

Exact H3 prompt
A 5-second 2D hand-drawn monochrome animation featuring the stylized character from <Picture 1> on a clean white background. In a medium full-body framing, the character descends lightly from the air, touches down onto an invisible ground plane, and settles into an alert fighting stance. Her living creature hair from <Picture 1> ripples and flexes its jagged jaws menacingly behind her. The character keeps her mouth closed and expression focused throughout. The camera remains static in a full-body view. Subtle swooshes and cloth rustle accompany the movement; no dialogue or vocalization.
#24 · ref2va-sample-2026083182-000023

A1 · 5s

eligiblecompletedraw rejecteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.88

The visual generation matches the input image and prompt well, but audible vocal sounds violate the no_speech requirement.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 2/4 · artifact 4/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately animates the characters from the reference image according to the prompt specifications without speech or visual flaws.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Make a 5-second anime video of the girl and guy from this image by the table, with her staying close to him while he nervously points at the paperwork, without any dialogue.

Exact H3 prompt
In the 2D anime illustration style of <Picture 1>, generate a continuous 5-second medium shot of the purple-haired girl with animal ears and the dark-haired young man in a vest from <Picture 1> standing together beside the desk. The girl gently rests her cheek against his face while holding onto his shirt. The young man blushes with wide, flustered eyes, his hand slightly trembling as he points toward the paper sheets on the table. Both characters keep their mouths closed with no speech or lip movement. The camera remains static at eye level with soft ambient indoor lighting. Quiet indoor room tone only.
#25 · ref2va-sample-2026083182-000024

A1 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video accurately reflects the reference character identity and executes the requested subtle combat pose, wind motion, camera push-in, and ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately depicts the character in a combat stance with matching anime styling, steady camera push-in, and compliant ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Make a 5-second anime video of the ninja girl from <Picture 1> in a ready combat stance as wind blows her hair.

Exact H3 prompt
In 2D anime style, a 5-second video of the ninja character from <Picture 1> poised in an alert battle stance in a dim forest clearing. Wind blows through her spiky blonde hair and rustles the surrounding leaves. Her sharp eyes narrow in focus while her mouth remains closed with no speech, no dialogue, no vocalization, no human voice, and no lip movement. The camera slowly pushes in on her face. Audio contains brisk wind and rustling foliage, with no speech, dialogue, or music.
#26 · ref2va-sample-2026083182-000025

A1 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The generated video accurately captures the character identity, removes the interface overlay, and renders natural animation consistent with the prompt instructions.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.9

The video faithfully animates the character and background from Picture 1 while successfully removing all UI elements.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create an 8-second anime-style animation of the blonde character wearing an apron from the screenshot. Show him standing in the culinary kitchen, cheerfully adjusting his apron and looking around the cooking station without any game interface on screen.

Exact H3 prompt
An 8-second 2D anime animation featuring the blonde character from <Picture 1> in the kitchen without any game UI. In a continuous medium shot, he stands with hands on hips, smiling. He opens his eyes, adjusts his apron bow, and glances across the prep counter with a cheerful nod. His mouth remains closed and at rest with no speech. Ambient kitchen hum and soft fabric rustle only, with no background music.
#27 · ref2va-sample-2026083182-000026

A1 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The generated video accurately translates the characters and scene from the reference image into subtle motion with matching ambient mechanical sound and no speech.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated clip faithfully animates the characters from the reference image in the requested setting with appropriate idle motion and compliant ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create an 8-second stylized anime clip of the characters from <Picture 1> sitting in the alleyway. The grey-haired boy and blue-haired girl watch their small round robot companion affectionately as it gently tilts and idles between them. Keep the scene entirely non-verbal with closed mouths, subtle smiles, gentle gestures, and ambient city/mechanical sound.

Exact H3 prompt
In an 8-second stylized anime sequence, generate the grey-haired young man, the blue-haired girl, and the rotund robotic companion from <Picture 1> sitting together against the tiled wall. The boy on the left and the girl on the right crouch beside the robot, watching it with warm, subtle smiles. The robot gently bobs and tilts side-to-side on its base with quiet mechanical twitches, and the girl reaches down slightly to pat its side while the boy gives a calm, approving nod. All mouths remain closed with no dialogue or vocalizations. The camera holds a steady medium shot with a very slow, subtle forward push-in. The audio consists of gentle urban alley ambience and soft mechanical clicks and whirs.
#28 · ref2va-sample-2026083182-000027

A1 · 8s

eligiblecompletedraw rejecteddelivery rejected

References

<Picture 1>
<Picture 1>subject_identity

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.88

The character identity and aesthetic are preserved, but the video suffers from severe transition glitches and an initial black background before rendering the environment.

task 3/4 · preserve 3/4 · identity 3/4 · motion 2/4 · audio 4/4 · artifact 2/4

failure codes: visible_artifacts, other

Delivery output

same audio as raw

Delivery QA

REJECTED · confidence 0.82

The character identity and target setting are maintained, but an abrupt transition and background change mid-clip disrupt temporal coherence.

task 3/4 · preserve 3/4 · identity 3/4 · motion 2/4 · audio 4/4 · artifact 2/4

failure codes: visible_artifacts

visual salvage: false

Natural request

Create an 8-second animated video of the anime girl hugging a pillow from Picture 1 relaxing in a cozy tatami room, gently swaying and resting happily without speaking.

Exact H3 prompt
In a clean 2D anime aesthetic with warm ambient lighting, the young woman with wavy strawberry-blonde hair hugging a soft lavender pillow from <Picture 1> reclines comfortably in a sunlit traditional Japanese tatami room. A medium shot frames her as she gently sways side to side with a content, smiling expression. Her eyes remain softly closed in relaxed bliss, and her mouth stays naturally at rest without any speaking or lip motion. Gentle breezes cause subtle strands of her hair to drift slightly. Soft ambient room tone and distant outdoor rustling leaves accompany the scene, with no spoken dialogue.
#29 · ref2va-sample-2026083182-000028

A3 · 10s

capability_canarycompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity
<Picture 2>
<Picture 2>subject_identity
<Picture 3>
<Picture 3>environment

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.85

The generated video accurately incorporates both subject identities and the environment reference, following the specified silent non-verbal interaction cleanly.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

silent by original contract

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately reflects all reference images and requested non-verbal interaction in the specified environment without notable artifacts.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 10-second silent video showing the woman from Picture 1 and the woman from Picture 2 browsing bookshelves together inside the bookstore shown in Picture 3. They interact quietly and non-verbally, examining a book and exchanging friendly smiles without speaking.

Exact H3 prompt
A realistic 10-second silent video set in the bookstore from <Picture 3>. The woman from <Picture 1> stands beside the woman from <Picture 2> in front of tall wooden bookshelves filled with books. The woman from <Picture 1> gently slides a hardcover book off a wooden shelf, opens it, and tilts it toward the woman from <Picture 2>. The woman from <Picture 2> looks over at the pages, nods approvingly, and smiles. Both women maintain closed mouths with no spoken dialogue or vocalization. The camera slowly dollies sideways along the aisle under warm, cozy bookstore lighting.
#30 · ref2va-sample-2026083182-000029

A3 · 12s

capability_canarycompletedraw rejecteddelivery review

References

<Picture 1>
<Picture 1>subject_identity
<Picture 2>
<Picture 2>subject_identity
<Picture 3>
<Picture 3>environment

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.92

The visual generation effectively places both subjects into the target environment, but contains audible invented speech violating the explicit no-speech requirement.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

silent by original contract

Delivery QA

REVIEW · confidence 0.82

The multi-subject composition successfully combines the requested subjects and beach environment with high identity fidelity and natural non-verbal interaction throughout the duration.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: false

Natural request

Create a 12-second silent video showing the woman from Picture 1 and the woman from Picture 2 walking together along the sandy beach from Picture 3, interacting non-verbally.

Exact H3 prompt
A 12-second silent video set on the sunny beach from <Picture 3>. The woman from <Picture 1> and the woman from <Picture 2> walk side by side across the sand past the surfboards. They look at each other, smile gently, and gesture toward the ocean. Both keep their mouths closed or naturally at rest, with no lip movement, no dialogue, no vocalization, and no human voice.
#31 · ref2va-sample-2026083182-000030

A3 · 5s

capability_canarycompletedraw rejecteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity
<Picture 2>
<Picture 2>subject_identity
<Picture 3>
<Picture 3>environment

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.95

While visual subject identities and the environment were integrated reasonably well, the output violates the required silent/no-speech constraints by generating unintelligible spoken audio.

task 3/4 · preserve 3/4 · identity 4/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

silent by original contract

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately incorporates all visual references, maintains character identities and environment details, and adheres to the motion and audio constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a silent 5-second clip set inside the antique bookstore from <Picture 3>. The woman from <Picture 1> and the sailor from <Picture 2> stand next to each other browsing the vintage bookshelf, exchanging a polite glance and nod without any speech.

Exact H3 prompt
A 5-second silent video set in the warm antique bookstore environment of <Picture 3>. The woman from <Picture 1> in her patterned pink gown and the sailor from <Picture 2> in full white uniform stand together in front of the vintage bookshelf. The woman gently reaches out toward a book on the shelf while holding her handbag, and the sailor turns slightly to offer a polite, friendly nod. Both subjects keep their mouths closed naturally with no speaking or lip motion throughout the entire shot.
#32 · ref2va-sample-2026083182-000031

A3 · 8s

capability_canarycompletedraw rejecteddelivery accepted

References

<Picture 1>
<Picture 1>subject_identity
<Picture 2>
<Picture 2>subject_identity
<Picture 3>
<Picture 3>environment

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.88

Visual identities, clothing, environment, and motion execution are well preserved, but the audio policy was violated by including background music instead of complete silence.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 2/4 · artifact 4/4

failure codes: audio_missing_or_changed

Delivery output

silent by original contract

Delivery QA

ACCEPTED · confidence 0.88

The video accurately combines the two target identities into the specified boardwalk environment with smooth motion and adherence to audio/speech policies.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create an 8-second silent video of the woman from Picture 1 and the man from Picture 2 walking together along the boardwalk from Picture 3, pausing to look around without speaking.

Exact H3 prompt
An 8-second silent, realistic shot on the wooden boardwalk from <Picture 3>. The woman from <Picture 1> and the man from <Picture 2> walk side-by-side down the walkway. Both pause and look around at the cabins. The camera tracks backward smoothly at eye level. There is no speech, no dialogue, no vocalization, no human voice, and no lip movement; their mouths remain closed and naturally at rest. Audio is silent.
#33 · ref2va-sample-2026083182-000032

A4 · 10s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video accurately presents the reference sphere on a wooden pedestal with a smooth orbiting camera move and clean studio presentation.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated video successfully portrays the reference sphere on a wooden pedestal with smooth orbital camera movement and clean studio presentation.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Make a clean product showcase video featuring the white sphere from the picture resting on a minimalist wooden pedestal with a smooth orbiting camera.

Exact H3 prompt
A clean studio product showcase video of the white sphere from <Picture 1> resting on a polished light-oak wooden pedestal against a soft neutral background. The camera performs a smooth, slow semi-circular orbit around the sphere, accentuating its matte surface under diffused lighting. Faint clean studio room tone. No speech.
#34 · ref2va-sample-2026083182-000033

A4 · 10s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video accurately reconstructs the reference object in a clean macro orbital pan, matching prompt constraints and audio policies.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The video executes the macro orbit showcase effectively, maintaining high identity fidelity to the reference image with clean camera movement.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a clean 10-second macro showcase video of the small pale sphere from <Picture 1> resting on the checkerboard sheet upon the wooden table. The camera slowly arcs around the sphere in a smooth cinematic orbit while the ball remains resting stably with subtle lighting shifts across its surface. No speech, with soft ambient room tone.

Exact H3 prompt
A clean 10-second cinematic macro showcase video focusing on the pale matte spherical ball from <Picture 1> resting on the black-and-white checkerboard pattern sheet on a light wood table. The camera executes a slow, steady orbital pan around the stationary sphere at a close macro perspective, capturing soft natural light reflections and fine shadows shifting across its curved surface and the grid beneath. No spoken dialogue; clean, subtle indoor ambient room tone only.
#35 · ref2va-sample-2026083182-000034

A4 · 10s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.86

The generated video accurately depicts the reference sphere rolling smoothly along a clean surface with consistent visual quality and fitting Foley audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.85

The model successfully generated a smooth close-up shot of the referenced blue sphere rolling across the table, satisfying all visual and audio criteria.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 10-second close-up video showcasing the light blue sphere from Picture 1 rolling smoothly across a clean tabletop surface.

Exact H3 prompt
A smooth, light blue sphere from <Picture 1> rolls slowly across a clean, neutral-toned tabletop in a well-lit studio setting. The camera tracks along at a low angle in a continuous close-up shot as the sphere rotates smoothly across the surface. Soft natural room reflections slide across the ball's curved exterior. The sound consists of a gentle, smooth rolling contact sound against the wooden table surface and subtle room tone; no speech or music.
#36 · ref2va-sample-2026083182-000035

A4 · 12s

eligiblecompletedraw rejecteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.82

The purple sphere identity and studio setting are well transferred, but an unnatural table surface deformation appears mid-video.

task 4/4 · preserve 4/4 · identity 4/4 · motion 3/4 · audio 4/4 · artifact 3/4

failure codes: visible_artifacts

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.85

The model successfully creates a studio showcase of the referenced purple sphere rolling across a polished wooden tabletop with high fidelity and compliant audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a clean studio showcase video of the purple sphere from the image rolling smoothly across a polished wooden tabletop.

Exact H3 prompt
A 12-second studio product showcase video featuring the purple spherical object from <Picture 1>. The purple sphere slowly rolls across a smooth, polished light oak tabletop in a brightly lit studio setting. The camera performs a smooth, close-up tracking shot, keeping the textured purple sphere in crisp focus while shallow depth of field softly blurs the warm minimalist background. The motion is calm and continuous, accompanied by subtle wooden contact rolling sounds and quiet studio ambience. No speech.
#37 · ref2va-sample-2026083182-000036

A4 · 12s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The generated video effectively demonstrates a clean commercial orbital shot around the reference folding cot while preserving object identity and complying with audio requirements.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.85

The generated video accurately replicates the folding cot from the reference image with smooth orbital camera movement and clean background audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a clean 12-second commercial showcase video of the folding camping cot from the picture, smoothly orbiting around the product in a minimalist studio environment to highlight its frame design and padded gray cushion.

Exact H3 prompt
A smooth commercial product showcase of the folding camping cot from <Picture 1>. The cot rests in the center of a spacious minimalist studio with polished light flooring and soft diffused daylight. The camera performs a slow, elegant 360-degree orbital tracking shot around the cot, emphasizing its sturdy black metal folding frame, ergonomic headrest contour, and dark gray padded fabric. The lighting creates gentle highlights along the tubular metal joints and fabric seams. Silent ambient studio atmosphere with no spoken dialogue.
#38 · ref2va-sample-2026083182-000037

A4 · 5s

eligiblecompletedraw accepteddelivery rejected

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video accurately replicates the object identity from the reference image onto the requested wooden surface with smooth motion and appropriate studio lighting.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

REJECTED · confidence 0.88

The model successfully places the referenced textured yellow sphere onto a wooden table with appropriate lighting and subtle motion consistent with the prompt.

task 3/4 · preserve 4/4 · identity 4/4 · motion 3/4 · audio 3/4 · artifact 4/4

failure codes: none

visual salvage: false

Natural request

Create a 5-second close-up video of the yellow textured sphere from <Picture 1> gently rolling across a clean wooden surface in bright studio lighting.

Exact H3 prompt
A close-up studio shot of the yellow textured sphere from <Picture 1> on a smooth wooden tabletop. The sphere slowly rolls across the frame in a smooth, steady motion under warm diffuse lighting, casting a soft moving shadow, while the camera tracks its path until it gently rests. No speech, accompanied only by subtle rolling Foley and quiet room tone.
#39 · ref2va-sample-2026083182-000038

A4 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The showcase video accurately preserves the object identity and tape markings while executing the requested camera move and workbench environment.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The model successfully creates a clean showcase video featuring the referenced wrench on a wooden surface with consistent identity and appropriate cinematic motion.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a smooth 5-second close-up product showcase video of the adjustable wrench from Picture 1 resting on a clean wooden workbench under studio lighting.

Exact H3 prompt
A 5-second cinematic product showcase shot of the adjustable wrench from <Picture 1> resting on a wooden workbench. The camera performs a smooth, low-angle orbital arc around the tool under warm studio lighting, accentuating its metallic head and blue handle with white bands. Quiet indoor room tone; no speech; no background music.
#40 · ref2va-sample-2026083182-000039

A4 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.86

The video successfully showcases the reference tin with smooth orbital motion on a wooden table, matching the prompt requirements and speech policy.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The video successfully showcases the reference tin box with a smooth orbit on a wooden table, adhering to all audio and visual constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Showcase the round cartoon pig tin from <Picture 1> with a slow, smooth 360-degree camera orbit on a clean wooden table.

Exact H3 prompt
A close-up commercial product showcase of the yellow round cartoon pig tin box from <Picture 1> resting on a polished wooden tabletop under soft studio lighting. The camera glides in a slow, smooth orbital pan around the tin to showcase its circular design and colorful pig illustrations from multiple angles. No dialogue or human voices are present, with only quiet ambient studio room tone.
#41 · ref2va-sample-2026083182-000040

A4 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.95

The generated video accurately features the reference fork in a clean kitchen setting with smooth orbital camera motion matching all specifications.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.85

The generated video accurately places the referenced fork into the requested modern kitchen scene with smooth camera motion and strong visual fidelity.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a clean 5-second product showcase video featuring the fork from Picture 1 resting on a modern kitchen countertop as the camera smoothly orbits around it.

Exact H3 prompt
A clean 5-second cinematic product showcase of the fork from <Picture 1> resting horizontally on a light wooden countertop in a brightly lit modern kitchen. The camera executes a slow, steady orbital arc around the fork, highlighting its purple tines and wrapped handle under soft diffused natural lighting. Silent scene with subtle ambient room tone; no spoken dialogue.
#42 · ref2va-sample-2026083182-000041

A4 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.86

The generated video accurately incorporates the reference object, performing the requested manipulation sequence smoothly with good visual fidelity and no speech violations.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately incorporates the reference pliers and executes the requested interaction sequence smoothly on a workbench setting without speech.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Can you create an 8-second video featuring the combination pliers from this image being picked up from a clean workbench by a pair of hands wearing work gloves, opening and closing the jaws smoothly, and then setting them back down?

Exact H3 prompt
Generate an 8-second video showcasing the combination pliers from <Picture 1>. On a clean wooden workbench in a brightly lit workshop, a person wearing grey work gloves reaches into the frame, picks up the pliers from <Picture 1>, opens and closes the metallic jaws twice to test their grip, rotates them slightly to show the orange-and-black handles, and gently places them back down on the surface. The camera holds a steady, crisp close-up shot with subtle depth of field. No spoken dialogue; ambient workshop room tone and realistic metallic handling Foley.
#43 · ref2va-sample-2026083182-000042

A4 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The video faithfully reproduces the specified tool in a clean workshop setting with fluid camera motion and consistent lighting.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated video successfully showcases the reference wrench in a cinematic workshop setting with smooth camera motion and faithful object details.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a smooth 8-second cinematic close-up video showcasing the adjustable wrench with the red handle from Picture 1 resting on a clean wooden workshop table, with the camera slowly panning around it to highlight its metallic finish and details under soft lighting.

Exact H3 prompt
A smooth, cinematic 8-second close-up showcase of the adjustable wrench with the red handle shown in <Picture 1>. The wrench rests atop a clean wooden workbench under soft directional workshop lights. The camera performs a slow, continuous panning arc around the tool, catching crisp metallic highlights on its silver adjustable jaw and knurled screw, while detailing the matte red grip of its handle. Quiet workshop ambient tone with faint room presence; no spoken dialogue.
#44 · ref2va-sample-2026083182-000043

A4 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.86

The model accurately rendered the specified billiard ball rolling smoothly across a clean surface with consistent visual details and compliant audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.85

The generated video successfully portrays the specified billiard ball rolling smoothly across a clean white table with accurate object details and high visual fidelity.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Can you generate a clean video of the number 15 pool ball from the picture rolling smoothly across a white tabletop under soft studio lighting?

Exact H3 prompt
Generate an 8-second realistic video of the burgundy striped number 15 billiard ball from <Picture 1> rolling smoothly across a clean white tabletop under bright, soft studio lighting. The camera tracks along in a close-up tracking shot as the ball rotates gracefully, reflecting soft overhead lights on its glossy polished surface. No spoken dialogue; only the faint, gentle rolling sound of a smooth billiard ball on a hard flat surface and subtle room tone.
#45 · ref2va-sample-2026083182-000044

A4 · 10s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The model successfully creates a commercial showcase matching the reference product identity, environment, and motion requests.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The model successfully placed the referenced cosmetic tube into the requested commercial setting with smooth rotation and accurate visual fidelity.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 10-second sleek product commercial featuring the cosmetic tube from the photo placed on an elegant pedestal in a bright studio with smooth camera rotation.

Exact H3 prompt
A 10-second commercial showcase featuring the cosmetic tube from <Picture 1> standing upright on a minimalist pedestal in a bright studio. The camera performs a smooth 360-degree rotating orbit around the tube under warm diffused lighting, concluding with a gentle upward tilt. Subtle ambient room tone and soft water ripples; no music, no speech.
#46 · ref2va-sample-2026083182-000045

A4 · 10s

eligiblecompletedraw rejecteddelivery review

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.82

The toy car identity and showcase camera motions are rendered well, but the requested forward driving motion of the vehicle is missing.

task 3/4 · preserve 3/4 · identity 4/4 · motion 3/4 · audio 3/4 · artifact 4/4

failure codes: unrequested_change

Delivery output

same audio as raw

Delivery QA

REVIEW · confidence 0.72

The generated video successfully produces a sleek product showcase of the reference toy police car with smooth camera movement and consistent object fidelity.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

visual salvage: false

Natural request

Make a sleek 10-second product showcase video of the small toy police car from the photo gliding across a modern tabletop with smooth camera movement highlighting its details.

Exact H3 prompt
A 10-second commercial product showcase video featuring the small white toy police car from <Picture 1>. The toy car glides smoothly forward across a polished light-toned table in a bright, modern studio setting with soft reflections. The camera begins at a low three-quarter angle tracking alongside the toy car as its wheels roll smoothly, then smoothly arcs around the front to capture its sleek body lines, tinted windows, and miniature rooftop light bar. The car comes to a gentle stop in the center of the frame as the camera executes a slow push-in hero shot. No human characters; no spoken dialogue; subtle sounds of miniature wheels rolling on a smooth surface and clean room tone.
#47 · ref2va-sample-2026083182-000046

A4 · 10s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The model successfully creates a clean product showcase video of the specified object with proper orbital motion and lighting.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The model successfully isolated the object and rendered a clean studio showcase matching all prompt criteria.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a sleek 10-second studio product showcase video featuring the blue chalk cylinder from the image standing on a clean modern pedestal with smooth orbital camera movement and soft ambient lighting.

Exact H3 prompt
A smooth, cinematic 10-second commercial product showcase featuring the chalk cylinder from <Picture 1> standing upright on a minimalist matte studio pedestal. The camera performs a steady orbit around the object while soft directional lighting sweeps across its textured surface against a clean neutral background. Accompanied by quiet ambient studio room tone; no speech.
#48 · ref2va-sample-2026083182-000047

A4 · 12s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The generated video accurately presents the cosmetic tube in an elegant commercial studio environment with smooth macro camera motion and compliant ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.86

The model successfully creates a commercial shot around the reference cosmetic tube with clean lighting, accurate object geometry, and smooth camera motion.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a sleek 12-second beauty product commercial showcasing the cosmetic tube from this picture in an elegant studio setting with soft lighting and smooth camera movements.

Exact H3 prompt
A 12-second commercial showcasing the cosmetic tube from <Picture 1> in a minimalist luxury studio with warm diffuse lighting. The camera executes smooth macro orbital movements around the tube, highlighting its transparent body and matte black cap. Soft ambient room tone without music or human speech.
#49 · ref2va-sample-2026083182-000048

A4 · 12s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video accurately reproduces the novelty glasses from the reference image in a smooth, high-quality commercial studio showcase meeting all requirements.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated video successfully preserves the identity of the novelty glasses from the reference image and executes a clean commercial studio showcase.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 12-second commercial product showcase video featuring the novelty glasses from <Picture 1>. Show smooth cinematic camera movements rotating and gliding around the glasses on a sleek modern display surface in studio lighting, highlighting their unique red frame and textured round lenses, with clean ambient studio atmosphere and no speech.

Exact H3 prompt
A 12-second commercial product showcase video featuring the novelty glasses from <Picture 1>. In [Shot 1], the camera performs a smooth cinematic orbit around the glasses from <Picture 1> resting on a polished matte pedestal under soft directional studio lighting. In [Shot 2] at 00:06.000, the shot cuts to an elegant macro tracking pan highlighting the vibrant red frame, curved temple arms, and textured round lenses. No speech, vocalization, or dialogue is present, accompanied solely by subtle indoor studio room ambience with faint air resonance.
#50 · ref2va-sample-2026083182-000049

A4 · 5s

eligiblecompletedraw rejecteddelivery rejected

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.82

The generated video delivers high visual quality and smooth motion showcasing a vintage model car; however, the car model body style differs from the reference, and unrequested background music is present.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: target_identity_lost, unrequested_change

Delivery output

same audio as raw

Delivery QA

REJECTED · confidence 0.85

The generated video delivers a sleek rotating product showcase, but alters the reference vehicle geometry from an open convertible to a hardtop sedan and includes musical audio.

task 3/4 · preserve 2/4 · identity 2/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: target_identity_lost, unrequested_change

visual salvage: false

Natural request

Create a 5-second sleek commercial product showcase of the vintage toy car from <Picture 1> on a rotating studio pedestal with cinematic lighting.

Exact H3 prompt
A sleek 5-second macro commercial showcase featuring the miniature vintage toy car from <Picture 1>. The red toy car rests atop a smooth, dark reflective studio pedestal under soft rim lighting. The camera executes a slow, elegant arc around the vehicle as the pedestal gently rotates, showcasing the glossy paintwork and chrome details. Clean ambient studio room tone; no speech or music.
#51 · ref2va-sample-2026083182-000050

A4 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.85

The generated video satisfies all commercial macro prompt requirements, featuring smooth camera motion and faithful object identity.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video successfully portrays the target potato chip on dark slate with warm lighting and smooth rotation, adhering to all audio and duration constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a 5-second cinematic macro commercial shot of the potato chip from Picture 1 resting on an elegant dark slate studio tabletop, filmed with a smooth circular rotating camera movement under warm studio lighting.

Exact H3 prompt
A 5-second cinematic product commercial macro shot of the golden-yellow curved potato chip from <Picture 1>. The crisp rests delicately on a clean dark slate surface under warm studio spotlighting. The camera performs a smooth, continuous low-angle circular orbit around the potato chip, capturing its crisp texture, subtle seasoning flecks, and delicate edges in sharp focus with soft background bokeh. No spoken dialogue or background music; only faint, quiet studio ambient room tone.
#52 · ref2va-sample-2026083182-000051

A4 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The model faithfully reconstructs the product identity from the reference image in a polished commercial showcase setting with smooth camera orbiting.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The model accurately recreated the hand cream tube identity from the reference image in a clean commercial tabletop setting with smooth motion and no defects.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a smooth 5-second product showcase video of the hand cream tube from the picture resting on a clean wooden table with soft daylight and subtle botanical accents.

Exact H3 prompt
A 5-second product commercial featuring the hand cream tube from <Picture 1> resting on a light natural wood tabletop alongside subtle green botanical leaves. The camera smoothly orbits the tube under soft diffuse daylight, highlighting its floral print and matte finish. No speech; quiet ambient room tone with soft air flow.
#53 · ref2va-sample-2026083182-000052

A4 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.85

The model accurately reconstructed the product in the requested commercial studio environment with smooth motion and appropriate background sound.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The video successfully showcases the product in a pastel aesthetic with smooth camera motion and accurate identity preservation.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a smooth 5-second commercial product showcase video featuring the hand cream tube from <Picture 1>, placing it in an elegant pastel-toned studio setting with soft lighting and a gentle camera orbit.

Exact H3 prompt
A cinematic commercial product shot showcasing the hand cream tube from <Picture 1>. The tube rests elegantly on a clean, soft pastel surface bathed in warm, diffused morning sunlight with subtle soft-focus background botanicals. The camera slowly orbits and glides closely around the tube, highlighting its packaging details and peach illustration. No speech; soft ambient studio tone with subtle soothing atmospheric whoosh.
#54 · ref2va-sample-2026083182-000053

A4 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.85

The generated video successfully showcases the reference cosmetic product in an elegant studio setup with smooth orbiting camera movement and clean execution.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.85

The generated video successfully isolates the cosmetic stick from the reference image, placing it on a marble pedestal in an elegant studio with smooth orbit motion.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create an 8-second commercial product showcase video featuring the pink rectangular cosmetic stick from <Picture 1>. Set it on a polished marble pedestal inside an elegant studio with soft lighting, with the camera smoothly orbiting around the product.

Exact H3 prompt
A sleek 8-second commercial product showcase of the pink rectangular cosmetic stick from <Picture 1>. The product rests upright on a polished white marble pedestal with soft specular reflections in an elegant, softly lit cosmetic studio with pastel pink and warm neutral gradients in the blurred background. The camera performs a smooth, continuous semi-orbital pan around the cosmetic stick, highlighting its matte pink texture and clean edges under diffuse studio key lights and gentle rim lighting. No spoken dialogue; soft atmospheric studio tone.
#55 · ref2va-sample-2026083182-000054

A4 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.9

The model successfully generated a commercial product showcase featuring the reference novelty glasses on a pedestal with steady camera motion and appropriate ambient sound.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately presents the pink novelty glasses in a studio setting with stable motion and high fidelity to the reference image.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a smooth 8-second commercial product video showcasing these pink flamingo party glasses on a modern studio table.

Exact H3 prompt
A clean 8-second commercial product showcase of the pink novelty flamingo eyeglasses from <Picture 1>. The glasses sit centered on a smooth pastel studio pedestal under soft, diffused lighting. The camera executes a slow, steady 360-degree orbital rotation around the glasses, highlighting their sculpted flamingo silhouette and glossy plastic finish. Clean ambient studio room tone, no speech, no music.
#56 · ref2va-sample-2026083182-000055

A4 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Picture 1>
<Picture 1>object_identity

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated video accurately presents the ridged chip from the reference in a high-quality studio showcase with smooth motion and compliant audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The model successfully creates a commercial showcase of the referenced potato chip floating and rotating under studio lighting.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Create a sleek 8-second commercial showcase video of the ridged potato chip from <Picture 1> rotating in slow motion under studio lighting.

Exact H3 prompt
A sleek 8-second commercial product showcase of the ridged potato chip shown in <Picture 1>. The golden-yellow crinkle-cut potato chip floats and rotates gracefully in mid-air in slow motion against a dark, minimalist studio backdrop with soft dramatic rim lighting. The camera slowly orbits and zooms in to capture the crisp texture and wavy ridges. As it turns, fine salt crystals subtly glisten across its surface before it lands softly onto a clean dark slate surface. Crisp, gentle whoosh and subtle crunchy ambient studio Foley; no spoken dialogue.
#57 · ref2va-sample-2026083182-000056

D4 · 10s

eligiblecompletedraw accepteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated continuation accurately completes the requested sequence of closing the filter assembly and pressing the panel button with consistent visuals and synchronized sound effects.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.95

The generated video smoothly continues the scene from the reference video, performing all requested interactions with consistent identity and natural foley audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the video from where it ends: have the person close the lid of the coffee machine's filter basket, swing the compartment shut, and select a brew setting on the digital control panel.

Exact H3 prompt
Continue the scene from the final frame of <Video 1>. The person presses the top lid of the filter compartment back down securely, swings the swing-out filter door fully closed into the coffee maker body, and then taps the start/control button on the upper silver panel. Natural mechanical clicks, plastic latching sounds, and subtle electronic beeps are heard with no spoken dialogue.
#58 · ref2va-sample-2026083182-000057

D4 · 10s

eligiblecompletedraw rejecteddelivery rejected

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.92

The continuation visually executes the kettle pouring action into the mug, but fails due to unrequested spoken dialogue violating the no-speech policy.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: unexpected_speech

Delivery output

speech removed after raw failure

Delivery QA

REJECTED · confidence 0.82

The continuation smoothly executes the requested action of pouring hot water into the mug, though the kettle material shifted from clear glass to stainless steel.

task 4/4 · preserve 3/4 · identity 3/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: target_identity_lost

visual salvage: false

Natural request

Continue the video from the kitchen stove, showing hot water from the glass kettle being poured into the mug next to it with realistic Foley and no spoken dialogue.

Exact H3 prompt
Continuing immediately from the end of <Video 1>, a hand reaches in to lift the glass kettle from the gas stove and pours steaming hot water into the ceramic mug sitting on the white tiled counter. Ambient kitchen sounds and the distinct trickling sound of hot water pouring are audible; no dialogue or music.
#59 · ref2va-sample-2026083182-000058

D4 · 10s

eligiblecompletedraw accepteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The continuation smoothly extends the scene, perfectly performing the requested action of adding green onions and tossing the salad with matching Foley sound and zero speech.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated continuation follows the prompt accurately, maintaining seamless visual continuity and appropriate sound effects.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the cooking video from <Video 1>. Have the chef add the chopped green onions from the small dish into the bowl and toss the salad together.

Exact H3 prompt
Continue seamlessly from <Video 1> for 10 seconds. From the exact final state where the chef finishes adding pepper, have the chef pick up the small dish of chopped green onions, tip them into the white bowl, and use salad spoons to gently toss and mix the salad ingredients together. Maintain the exact kitchen countertop setting, static medium-close camera framing, and lighting from <Video 1>. No speech; include natural Foley sounds of ceramic dishes, utensils, and food mixing.
#60 · ref2va-sample-2026083182-000059

D4 · 10s

eligiblecompletedraw accepteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The continuation smoothly extends the scene, plating additional breakfast items from the same perspective with realistic motion and appropriate ambient Foley audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated continuation seamlessly extends the source video with accurate serving actions, consistent tabletop environment, and appropriate ambient audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the video seamlessly from where it left off, showing the person serving fried potatoes and baked beans onto the breakfast plate, keeping the same first-person perspective and tabletop setting without any talking.

Exact H3 prompt
Continue seamlessly from the end of <Video 1> from the same overhead first-person perspective at the breakfast table. Using tongs, the person transfers crispy fried potato slices from the cast-iron skillet and a spoonful of baked beans onto the plate alongside the egg, bacon, sausage, and cooked tomato. The camera maintains a stable high-angle shot of the patterned tablecloth and breakfast spread. No spoken dialogue; natural clinking of cutlery, tongs, and quiet ambient room sounds only.
#61 · ref2va-sample-2026083182-000060

D4 · 10s

eligiblecompletedraw accepteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.86

The generated video seamlessly continues the source clip, completing the garlic separation and inspection with matching visual continuity and appropriate sound effects without speech.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated continuation accurately maintains visual continuity, follows the requested actions logically, and complies with the audio requirements.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the clip seamlessly from where it left off, showing the person finish separating the garlic cloves from the bulb and inspecting one in their hand.

Exact H3 prompt
Continue seamlessly from the final frame of <Video 1> in a continuous high-angle close-up shot. The person from <Video 1> finishes separating the loosened garlic cloves from the central stalk using their fingers, places the small pry tool down onto the dry straw ground, and holds up a single garlic clove between their thumb and index finger to inspect its skin. Maintain identical outdoor daylight, clothing, and background. No spoken dialogue; include only subtle outdoor ambience and the gentle crisp rustle of garlic husks and straw.
#62 · ref2va-sample-2026083182-000061

D4 · 12s

eligiblecompletedraw accepteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.9

The generated continuation smoothly follows the context video with accurate character continuity, prompt alignment, and natural kitchen Foley.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated continuation faithfully preserves the subject's identity, wardrobe, and environment while accurately executing all requested sequence actions and audio constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue from <Video 1> as the woman places the bowl in the fridge, closes the refrigerator door, and walks back over to the kitchen counter to continue preparing food, with no talking.

Exact H3 prompt
Continue directly from the end of <Video 1>. The woman from <Video 1> places the small red dish onto a refrigerator shelf, pushes the refrigerator door shut, and walks back over to the kitchen counter. She looks over the ingredients spread across the island counter and picks up a small bowl, keeping her mouth closed with no spoken dialogue. Gentle indoor kitchen room tone, refrigerator door closing thud, and soft countertop handling sounds accompany the motion.
#63 · ref2va-sample-2026083182-000062

D4 · 12s

eligiblecompletedraw accepteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated continuation follows the context smoothly, demonstrating the requested sewing action with consistent visuals and compliant audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 3/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated continuation naturally proceeds from the reference context, depicting the stitching action with consistent objects, hands, and environment while complying with all audio constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the video by showing the person beginning to sew the knitted pouch onto the metal clasp frame using needle and thread.

Exact H3 prompt
Continue seamlessly from <Video 1> for 12 seconds. Starting from the final frame where the hands hold the top edge of the grey knitted bag against the metal clasp frame, the person takes a needle threaded with black cotton yarn and begins carefully stitching the crocheted edge into the inner holes of the metal frame along the rim. The camera maintains the static top-down close-up view over the tablecloth. No spoken dialogue; ambient crafting room tone and subtle textile handling Foley only.
#64 · ref2va-sample-2026083182-000063

D4 · 12s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.88

Visual continuation and hair styling actions align well with the reference context, but the generation violates the no-speech constraint with corrupted vocal audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 2/4 · artifact 4/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.88

The generated continuation smoothly proceeds from the reference context, accurately performing the requested hair-rolling task while preserving visual consistency and following audio constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue directly from the end of the video, showing the woman wrapping and securing the tail of her braid with a roller in the same bathroom setting, keeping her mouth closed and demonstrating non-verbally with ambient room audio.

Exact H3 prompt
Continue from the final frame of <Video 1> for 12 seconds. The woman in the plaid shirt smoothly wraps the end of her right braid around a roller, rolling the section upward toward the braid base and securing it neatly with a clamp. She keeps her mouth closed and naturally at rest throughout, demonstrating non-verbally without speaking. The camera maintains the steady medium framing in the same bathroom background. No spoken dialogue; subtle hair rustling, gentle plastic clamp clicking, and quiet indoor ambient room tone only.
#65 · ref2va-sample-2026083182-000064

D4 · 5s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.92

Visual continuation and cloth wiping action are executed well, but the output violates the no-speech constraint by generating unintelligible spoken audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 2/4 · artifact 4/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.92

The generated video executes the continuation accurately with seamless temporal consistency and proper adherence to the wipe action prompt.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the video seamlessly from the end: keep the camera on the table as the person's hands continue handling the small doll and use the cloth to gently clean its face.

Exact H3 prompt
Continue seamlessly from the end of <Video 1> for 5 seconds. The person's hands continue holding the small blonde doll in the pink dress on the dark tabletop, gently tilting it and picking up the white cloth to wipe its face. Handheld close-up camera angle remains steady, capturing natural tabletop handling Foley and room tone without spoken dialogue.
#66 · ref2va-sample-2026083182-000065

D4 · 5s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.88

The generated video fails the continuation boundary requirement by resetting to the beginning shot instead of advancing from the final frame.

task 2/4 · preserve 3/4 · identity 3/4 · motion 2/4 · audio 3/4 · artifact 3/4

failure codes: continuation_boundary_failed

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.85

The generated video successfully extends the context video, preserving lighting, subject details, and subtle camera motion.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the video of the two glowing, decorative candle holders flickering softly in the dark room as the camera slowly drifts closer.

Exact H3 prompt
Continue seamlessly from the end of <Video 1> for 5 seconds. Maintain the dark room setting with the two decorative candle holders glowing and flickering softly with warm orange light. The camera continues a slow, subtle close-up drift around the patterned lamps as soft shadows dance in the background. No dialogue or human vocalization; quiet ambient room tone only.
#67 · ref2va-sample-2026083182-000066

D4 · 5s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.9

Visual continuation is smoothly executed with the cloth wiping motion, but the audio track contains unexpected and unintelligible vocalization.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.85

The continuation smoothly extends the scene, preserving identity, camera angle, and context while performing the requested wiping action.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the clip by having the hand pick up the towel and wipe down the stainless steel stove surface.

Exact H3 prompt
Seamlessly continue from the ending of <Video 1>. The hand picks up the green and white striped cloth and gently wipes the stainless steel surface beside the gas stove burner. The camera maintains a steady close-up angle over the dark countertop and kitchen backsplash. Non-verbal action only with no speech, dialogue, or vocalization, accompanied by quiet kitchen room tone and soft cloth wiping sounds.
#68 · ref2va-sample-2026083182-000067

D4 · 5s

eligiblecompletedraw accepteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The continuation seamlessly extends the cutting action with high consistency in visual details and Foley audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.92

The generated video accurately continues the scene from the context reference, preserving character details, objects, and environment while rendering realistic cutting motion and matching audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the video as the hands use the scissors to cut along the traced line on the cardboard.

Exact H3 prompt
Continue seamlessly from the final frame of <Video 1> for 5 seconds. The hands with bright orange nail polish continue using the large metal scissors to cut steadily along the drawn pen line across the brown corrugated cardboard on the polka-dot table. Maintain the high-angle close-up framing, natural lighting, and surface setting. Audio includes Foley sounds of scissors slicing through thick cardboard and light paper handling, with no speech or music.
#69 · ref2va-sample-2026083182-000068

D4 · 5s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.85

The continuation completes the requested scooping action with appropriate cooking Foley, but introduces a noticeable jump cut at the boundary rather than a seamless frame transition.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 3/4 · artifact 4/4

failure codes: continuation_boundary_failed

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.88

The generated continuation accurately captures the requested action, maintaining visual and object consistency with natural motion and appropriate ambient sound.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue from the end of the video, showing the person finish scooping the remaining grilled mushrooms from the electric grill onto the white plate with the red spatula.

Exact H3 prompt
Continue seamlessly from the exact end frame of <Video 1>. The person continues using the red spatula to scoop the remaining grilled mushroom slices off the ridged electric grill and transfers them onto the white plate with the grilled onions. Maintain the same high-angle countertop framing, indoor kitchen lighting, and grill setup throughout. No spoken dialogue; natural cooking Foley including sizzling and spatula scraping.
#70 · ref2va-sample-2026083182-000069

D4 · 5s

eligiblecompletedraw rejecteddelivery rejected

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.92

The continuation fails boundary consistency by resetting to earlier video states, and violates the no-speech policy with unintelligible generated speech.

task 2/4 · preserve 3/4 · identity 3/4 · motion 2/4 · audio 2/4 · artifact 3/4

failure codes: continuation_boundary_failed, unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

REJECTED · confidence 0.88

The continuation fails boundary consistency by restarting the action from the beginning of the chart rather than resuming from the final frame.

task 2/4 · preserve 3/4 · identity 3/4 · motion 2/4 · audio 3/4 · artifact 3/4

failure codes: continuation_boundary_failed

visual salvage: false

Natural request

Continue the video from where it leaves off, showing the hand holding the green liquid vial over the phosphate color chart and slightly adjusting it between the color circles to compare the reading.

Exact H3 prompt
Continue seamlessly from the final frame of <Video 1>. From the fixed overhead top-down view, the hand holding the clear vial of green liquid gently shifts and rotates the vial over the printed green circles on the PO4 phosphate chart to compare the color value, before coming to a steady rest. The overhead framing, white background, bright diffuse lighting, and subject appearance remain perfectly consistent with the end of <Video 1>. No speech; faint plastic-on-paper friction and quiet room tone only.
#71 · ref2va-sample-2026083182-000070

D4 · 5s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.95

The visual continuation seamlessly tracks the sea lion and trainer, but the audio generates unintelligible spoken narration violating the no-speech constraint.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.92

The generated video smoothly continues the action from the reference clip with consistent identities, realistic motion, and adherence to the requested camera movement and audio constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the video from the final frame of <Video 1> for 5 more seconds. Show the trainer finishing adjusting the harness on the sea lion as the sea lion turns and waddles forward toward the open doorway near the pool, with no spoken dialogue.

Exact H3 prompt
Continue seamlessly from the final frame of <Video 1> for 5 seconds. The bald trainer in the pink shirt finishes securing the reflective harness around the sea lion from <Video 1>, giving it an encouraging pat on its flank. The sea lion turns its head forward and shifts its weight, waddling on its flippers toward the open doorway leading to the outdoor pool. The camera pans slightly right to follow the sea lion's forward movement. No spoken dialogue; mouth remains closed with ambient sounds of wet flippers slapping against the damp deck and water rippling nearby.
#72 · ref2va-sample-2026083182-000071

D4 · 5s

eligiblecompletedraw rejecteddelivery rejected

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.88

The continuation visually follows the prompt but fails strictly due to the generation of unexpected, partially unintelligible speech violating the no-speech requirement.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

REJECTED · confidence 0.82

The continuation maintains character identity and kitchen setting while remaining silent as requested, but suffers from a noticeable boundary discontinuity from the end of the context clip.

task 3/4 · preserve 3/4 · identity 3/4 · motion 2/4 · audio 4/4 · artifact 3/4

failure codes: continuation_boundary_failed

visual salvage: false

Natural request

Continue the video for 5 seconds as the woman finishes shucking the ear of corn in her hands and places it on the tray without speaking.

Exact H3 prompt
Continue seamlessly from the end of <Video 1> for 5 seconds: the woman in the kitchen continues shucking the ear of corn she is holding, stripping away the remaining husks and loose silks before placing the cleaned cob onto the baking tray, keeping her mouth closed at rest with no spoken dialogue; realistic rustling husks and ambient kitchen room tone only.
#73 · ref2va-sample-2026083182-000072

D4 · 8s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.88

The video continuation visually follows the folding action, but violates the no_speech requirement due to generated vocal speech.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.92

The generated continuation seamlessly extends the context video, accurately fulfilling the requested folding and smoothing actions while preserving subject and environmental consistency.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the video from where it leaves off: have the hands lay the Z-fold card flat on the table, fold it shut into its compact front-facing card position, and gently press the creases smooth.

Exact H3 prompt
Continue seamlessly from the ending of <Video 1>. The hands gently place the standing Z-fold card flat onto the wooden tabletop, fold the accordion panels inward until the card is fully closed into its compact rectangular form, and run a finger lightly along the folded edges to smooth the creases. Maintain the same overhead camera angle, lighting, and wooden surface. No speech or vocalization; subtle paper handling sounds and table contact Foley only.
#74 · ref2va-sample-2026083182-000073

D4 · 8s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.95

The visual continuation seamlessly follows the reference scene, but the output generates unrequested and garbled speech in violation of the no_speech requirement.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 2/4 · artifact 4/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.86

The generated continuation successfully executes the requested inspection action while preserving identities, setting consistency, and silence constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue directly from the end of the video. Have the two technicians inspect their work by stepping forward, opening the appliance door to check inside, and confirming the fit, without any talking.

Exact H3 prompt
Continue seamlessly from the final frame of <Video 1>. The two technicians in brown work shirts and khaki pants inspect the newly installed under-counter appliance. One technician steps forward, crouches down slightly, and pulls open the stainless steel door to check the interior rack and alignment, while the second technician observes from beside the counter. They exchange a brief approving nod. Keep both technicians' mouths closed and naturally at rest throughout, with no dialogue or vocalizations. Ambient indoor room tone and subtle metallic door handling sounds only.
#75 · ref2va-sample-2026083182-000074

D4 · 8s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.95

The continuation visually follows the character and action well, but fails strictly due to the generation of unexpected unintelligible speech violating the no_speech policy.

task 3/4 · preserve 3/4 · identity 4/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.88

The generated continuation accurately preserves the subject's identity, wardrobe, and actions from the reference video, following the requested motion and silent audio policies.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the video as the woman finishes blending her makeup across her face and checks her reflection in the mirror.

Exact H3 prompt
Continue seamlessly from the final frame of <Video 1>. The woman from <Video 1> in the olive jacket continues brushing her face with the pink makeup brush, blends the powder smoothly across her cheeks and jawline, and then lowers the brush slightly while tilting her head to inspect her reflection in the mirror. Her mouth remains naturally closed with no spoken dialogue, accompanied by quiet bedroom ambience and subtle soft brush swishes.
#76 · ref2va-sample-2026083182-000075

D4 · 8s

eligiblecompletedraw accepteddelivery rejected

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.92

The generated continuation faithfully preserves identity, environment, and motion continuity while fulfilling all prompt instructions.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

REJECTED · confidence 0.88

The continuation accurately adheres to the prompt, preserving subject identity, scene layout, and requested baking actions with minor object morphing on the cooling rack.

task 4/4 · preserve 4/4 · identity 4/4 · motion 3/4 · audio 4/4 · artifact 3/4

failure codes: none

visual salvage: false

Natural request

Continue the video as the baker finishes piping cream into the pastry, sets it down on the cooling rack, and picks up another puff.

Exact H3 prompt
Continue directly from the final frame of <Video 1> in the same static medium framing and kitchen setting. The woman shown in <Video 1>, wearing a black beret, navy shirt, dark overalls, and white gloves, finishes piping chocolate cream into the choux pastry, gently places the filled puff on the wire cooling rack on the wooden table, and picks up another unfilled pastry to begin piping filling into it. Her mouth remains closed throughout with no speech, accompanied only by quiet ambient kitchen room tone and subtle pastry handling sounds.
#77 · ref2va-sample-2026083182-000076

D4 · 8s

eligiblecompletedraw rejecteddelivery rejected

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.86

The requested action is performed accurately with appropriate Foley audio, but the camera position and scene layout change abruptly at the continuation boundary.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 4/4 · artifact 4/4

failure codes: continuation_boundary_failed, unrequested_change

Delivery output

same audio as raw

Delivery QA

REJECTED · confidence 0.82

The continuation accurately captures the garlic peeling action and audio policy, but introduces minor scene layout discontinuities.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 4/4 · artifact 4/4

failure codes: continuation_boundary_failed, unrequested_change

visual salvage: false

Natural request

Continue the video as the person separates the individual garlic cloves on the cutting board and begins peeling them.

Exact H3 prompt
Continue directly from the final frame of <Video 1> for 8 seconds: the person finishes separating the garlic bulb and begins using their fingers to peel the dry papery skins from the individual garlic cloves on the white cutting board next to the electric pressure cooker. The camera maintains its stationary medium-close angle. No dialogue or speech occurs; only the natural Foley sounds of peeling garlic skins and ambient room tone are heard.
#78 · ref2va-sample-2026083182-000077

D4 · 8s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.95

Visual continuation is preserved smoothly, but the presence of spoken dialogue directly violates the strict no_speech policy.

task 3/4 · preserve 3/4 · identity 3/4 · motion 3/4 · audio 2/4 · artifact 3/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.92

The generated video smoothly continues the unboxing sequence, maintaining identity, scene consistency, and requested prop interactions without artifacts or unprompted audio.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the unboxing video from the end of the clip. Have the presenter place the dongle back into the box and inspect the charging cable on the table without speaking.

Exact H3 prompt
Continue seamlessly from the end frame of <Video 1>. The woman in the floral blouse places the white adapter dongle into the open partition of the retail box on the white table. She reaches into the packaging to lift out the coiled white charging cable, inspecting the connectors under the studio lighting before resting it beside the smartphone. The camera maintains a fixed medium-close angle. The presenter keeps her mouth closed with no dialogue; only tactile prop handling sounds and quiet studio room tone are heard.
#79 · ref2va-sample-2026083182-000078

D4 · 5s

eligiblecompletedraw rejecteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

REJECTED · confidence 0.88

Visual continuation is consistent and executes the requested actions, but the generation fails due to unintended garbled spoken dialogue violating the no-speech constraint.

task 3/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 2/4 · artifact 4/4

failure codes: unexpected_speech, unintelligible_speech

Delivery output

speech removed after raw failure

Delivery QA

ACCEPTED · confidence 0.92

The generated continuation smoothly follows the reference video, executing the requested bar actions and garnish placement accurately while maintaining visual and identity consistency.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the video for another 5 seconds as the bartender finishes pouring, sets down the cocktail shaker on the counter, and garnishes the drink, with no spoken dialogue.

Exact H3 prompt
Continue seamlessly from the end of <Video 1> for 5 seconds. The mixologist in the light green shirt finishes pouring the drink into the martini glass, sets the metal shaker down on the bar counter, and places a garnish onto the rim of the glass. The host in the dark blazer stands beside him watching. The camera remains in a steady medium shot behind the bar. All characters keep their mouths closed with no spoken dialogue; include realistic clinking bar Foley and room ambience.
#80 · ref2va-sample-2026083182-000079

D4 · 8s

eligiblecompletedraw accepteddelivery accepted

References

<Video 1>continuation_context

Raw H3 output

Raw Gemini 3.7 QA

ACCEPTED · confidence 0.88

The generated video successfully continues the unboxing sequence with consistent product identity and proper handling motion.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

Delivery output

same audio as raw

Delivery QA

ACCEPTED · confidence 0.89

The continuation naturally completes the unboxing action, maintaining consistent object identity and handling motion while adhering to all prompt constraints.

task 4/4 · preserve 4/4 · identity 4/4 · motion 4/4 · audio 4/4 · artifact 4/4

failure codes: none

visual salvage: true

Natural request

Continue the unboxing video from the end of the clip, showing the hands sliding out the inner cardboard tray and removing the fitness band from its packaging.

Exact H3 prompt
Continue directly from the end of <Video 1> in the same close-up studio setting. The hands pull out the inner tray from the cardboard packaging, slide the contents into full view, and lift the packaged fitness band to display it clearly to the camera. Mouths remain closed and silent; clean studio ambience and gentle cardboard handling Foley only.