Gemini (edit-success protocol) passes refwarm ep019 on all three questions — object visible and coherent (VES), its sound present and coherent (AES), and audio/video in sync (AVC) — on 406 of 1,425 test clips. This page shows 50 of those 406, drawn at random (seed 0): it is a filtered sample of successes, not a random sample of the test set. For unfiltered examples see the 100-clip page. Baselines are shown on the same clips.
Full-test numbers for this direction: VES 40.1 / AES 77.9 / AVC 71.5 (GT ceiling 84.8 / 82.3 / 79.6); PSNR 19.22 · LPIPS 0.228 · FVD 316.7 · FAD 5.77 · LSD 11.85.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The owl is visually present and consistent throughout the clip.
AES Yes — Bird chirping sounds are audibly present and consistent.
AVC Yes — No checkable event; no observed mismatch as the owl does not move its beak.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The water flow is clearly visible and remains a consistent, continuous stream throughout the video.
AES Yes — A continuous rushing water sound is present and remains consistent without any unnatural dropouts.
AVC Yes — The continuous rushing sound perfectly matches the continuous visual flow of water; no mismatch observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The train is visually present and maintains a consistent appearance and trajectory as it moves away.
AES Yes — The train sound is audible and remains a coherent, consistent acoustic source throughout the clip.
AVC Yes — The continuous train sound matches the ongoing motion of the visible train, with no observed mismatch.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The chimpanzee is visually present, maintains a consistent appearance, and moves naturally through the scene until exiting.
AES Yes — Chimpanzee vocalizations and movement sounds are consistently audible and appropriate for the scene.
AVC Yes — The rustling sounds align with the visible movement through the vegetation, and the vocalizations are plausible for the animal's actions.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The bird is visually present and remains a coherent, consistent object throughout the clip.
AES Yes — The bird chirping sound is audibly present and remains a consistent acoustic source.
AVC Yes — The chirping sounds are consistent with the bird's presence and general movements; no clear audio-visual mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The keyboard is visually present and remains consistent in appearance and position across the entire clip.
AES Yes — The piano sound is clearly audible and maintains a consistent timbre and presence throughout.
AVC Yes — The woman's finger movements on the keys appear synchronized with the piano notes heard.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The clarinet is clearly visible and consistently held and played by the person from beginning to end.
AES Yes — The clarinet sound is clearly audible and consistent, matching the expected timbre of the instrument.
AVC Yes — The audible notes are synchronized with the visible finger movements and playing actions.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The bald man is visually present and maintains a consistent appearance and position throughout the entire clip.
AES Yes — The sound of the bass guitar, which is the expected audio source for the man's actions, is present and consistent.
AVC Yes — The visible plucking and fretting actions of the man on the bass guitar are synchronized with the audible bass notes.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The high-performance vehicle is visually present and maintains a consistent identity and trajectory across the entire clip.
AES Yes — The engine and tire squealing sounds are continuously audible and consistent with the target vehicle.
AVC Yes — The engine revs and tire squeals are synchronized with the visible drifting action of the car.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The train is visually present and maintains a consistent appearance and trajectory as it moves away.
AES Yes — The train's chugging sound is continuously present and consistent without unexplained dropouts.
AVC Yes — The continuous chugging sound aligns well with the ongoing movement of the train; no mismatch observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The dog is visually present and maintains a consistent appearance and motion throughout the clip.
AES Yes — The barks are consistent in timbre and appropriate for the visible dog.
AVC Yes — No checkable event; the dog's mouth is not clearly visible to confirm exact synchronization, but no clear mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The canine is visually present and maintains a consistent appearance and position throughout the clip.
AES Yes — The howling sound is audibly present and remains a consistent acoustic source throughout the clip.
AVC Yes — The howling sounds align appropriately with the dog's open mouth, with no observed mismatch.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The bells are clearly visible and swing consistently without any visual glitches or disappearances.
AES Yes — The ringing sound of the bells is present and consistent throughout the clip, matching the expected sound.
AVC Yes — The continuous ringing sound is synchronized with the continuous swinging motion of the bells.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The bird is visually present and remains a coherent, consistent object throughout the entire video.
AES Yes — The bird cooing sound is audibly present and remains consistent without any unnatural dropouts.
AVC Yes — The continuous cooing sound is appropriate for the visible bird, and no clear audio-visual mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The birds are clearly visible and maintain consistent identity and presence throughout the scene.
AES Yes — Appropriate bird calls are consistently audible throughout the clip without unnatural dropouts.
AVC Yes — The continuous chirping sounds align well with the general activity of the birds in the nest, with no observable mismatch.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The train, which is the source of the horn, is consistently visible and moves naturally throughout the clip.
AES Yes — The train horn is clearly audible, sounds appropriate for the visible train, and occurs naturally.
AVC Yes — The horn sounds while the train is present and moving; no specific visible action requires exact synchronization, and no mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The fingers of an individual are visually present and consistent throughout the clip.
AES Yes — The tapping sounds are audibly present and consistent throughout the clip.
AVC Yes — The tapping sounds are synchronized with the visible finger movements.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — Water is continuously visible as the sea and splashes throughout the entire clip.
AES Yes — The rushing and splashing sound of water is continuously audible and consistent with the scene.
AVC Yes — The continuous rushing sound matches the ongoing visual action of the boat moving through the water; no mismatch observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The vehicle is visually present and consistent across the entire clip.
AES Yes — The operational sound of the vehicle is continuous and consistent throughout.
AVC Yes — No checkable event; no observed mismatch.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The harp is visually present and consistent across the entire clip.
AES Yes — The harp sound is audibly present and consistent throughout the clip.
AVC Yes — The visible plucking of the harp strings is synchronized with the audible harp notes.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The aircraft's engines and spinning propellers are clearly visible and remain consistent throughout the camera pan.
AES Yes — A continuous and appropriate engine sound is present without any unnatural dropouts or changes.
AVC Yes — The continuous engine sound matches the continuous spinning of the propellers; no mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The water is consistently visible and its swirling motion is coherent across the entire clip.
AES Yes — The rushing water sound is continuously present and consistent with the visual action.
AVC Yes — The continuous sound of rushing water perfectly matches the ongoing visible swirling action with no observed mismatch.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The high-pitched percussion instruments (glockenspiels) are clearly visible and consistently played across the entire clip.
AES Yes — The corresponding high-pitched metallic percussion sounds are audibly present and consistent throughout.
AVC Yes — The visible mallet strikes on the instruments appear synchronized with the audible percussion notes; no clear mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The man is visually present and consistent throughout the clip.
AES Yes — The sound of the man playing the instrument is present and consistent.
AVC Yes — The man's plucking and sliding hand movements are synchronized with the guitar notes.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The canine is visually present and maintains a consistent identity and natural motion across the entire clip.
AES Yes — The barking sound is audibly present, consistent in timbre, and continues naturally throughout the clip.
AVC Yes — The barking sounds are synchronized with the dog's energetic movements; no clear audio-visual mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The harp is visually present and remains a consistent object throughout the clip.
AES Yes — The harp sound is present and remains a coherent acoustic source throughout.
AVC Yes — The woman's hand movements plucking the strings appear synchronized with the audible harp notes.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The engine is clearly visible and remains a coherent object throughout the camera pan.
AES Yes — The sound of the running engine is continuously audible without unnatural dropouts.
AVC Yes — No checkable discrete events; the continuous engine sound matches the visible running engine with no observed mismatch.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The crowd is visually present at the bottom of the screen and remains consistent throughout the clip.
AES Yes — The crowd cheering and clapping sound is continuously audible and consistent across the clip.
AVC Yes — No checkable event; no observed mismatch between the continuous crowd noise and the visible audience.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The percussion instruments are clearly visible and remain consistent throughout the entire clip.
AES Yes — The sounds of the percussion instruments are present and consistent with the ongoing performance.
AVC Yes — The visible mallet strikes on the instruments are synchronized with the audible notes.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The piano is visually present and remains a consistent object throughout the clip.
AES Yes — The piano sound is clearly audible and consistent across the entire clip.
AVC Yes — The visible key presses by the person's hands appear synchronized with the audible piano notes.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The water is clearly visible and flows consistently across the scene for the duration of the clip.
AES Yes — The rushing sound of water is continuously audible and consistent with the visual.
AVC Yes — The continuous rushing sound matches the continuous flow of the water; no mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The car engine components are visually present and consistent across the entire clip.
AES Yes — The sound of the running engine is continuously present and consistent without any dropouts.
AVC Yes — No checkable event; no observed mismatch between the continuous engine sound and the static engine visual.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The didgeridoo is visually present and maintains a consistent appearance and position throughout the clip.
AES Yes — The didgeridoo sound is audibly present and remains a consistent acoustic source throughout.
AVC Yes — The continuous visible action of playing matches the continuous sound; no mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The train is visually present and maintains its identity, structure, and trajectory consistently throughout the clip.
AES Yes — The train's chugging and hissing sounds are continuously present and remain a coherent acoustic source.
AVC Yes — The rhythmic chugging sounds are synchronized with the train's slow forward movement, with no observed mismatch.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — Guinea fowl is consistently visible throughout.
AES Yes — Guinea fowl calls are consistently audible.
AVC Yes — No checkable beak movements for calls; rustling matches visible motion.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The engine components are clearly visible and consistently moving throughout the entire clip.
AES Yes — The engine sound is continuously audible without any dropouts or unexplained changes.
AVC Yes — The continuous sound of the engine perfectly matches the continuous spinning of the visible pulleys.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The lion is continuously visible and coherent.
AES Yes — The lion roars are consistent in timbre and source identity.
AVC Yes — The audible roars synchronize with the lion's visible mouth movements and body heaves.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The aircraft engines are clearly visible and remain consistent in appearance and position for the entire duration.
AES Yes — The continuous sound of aircraft engines is present and remains consistent without any dropouts or unnatural changes.
AVC Yes — The continuous engine sound aligns with the visible engines; no discrete events to check, and no mismatch observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The harp is visually present and maintains a consistent appearance and position throughout the clip.
AES Yes — The harp music is continuously audible and consistent in timbre without any unnatural dropouts.
AVC Yes — The visible plucking of the harp strings synchronizes well with the audible notes.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The harp is clearly visible and consistent throughout the entire clip.
AES Yes — The sound of the harp is continuous and consistent.
AVC Yes — The hand movements plucking the strings synchronize well with the audible harp notes.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The power tool is visually present and remains a consistent, coherent object throughout the entire clip.
AES Yes — The sound of the power tool is audibly present and remains a consistent acoustic source without unexplained dropouts.
AVC Yes — The continuous sound of the tool operating matches the continuous visible action of the tool being used, with no observed mismatch.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The train is clearly visible and maintains its identity and motion throughout the clip.
AES Yes — The train's chugging sound is present and consistent for the duration of the clip.
AVC Yes — The rhythmic chugging sound aligns well with the visual movement of the train.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The chainsaw is clearly visible, identifiable, and maintains a consistent appearance and spatial relationship with the tree throughout the clip.
AES Yes — The chainsaw engine sound is continuously audible, coherent, and appropriate for the visible machinery without any unexplained dropouts.
AVC Yes — The continuous sound of the engine and cutting matches the ongoing visible action of the chainsaw cutting into the tree, with no observed mismatch.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The harp is visually present and remains temporally consistent across the entire clip.
AES Yes — The sound of the harp is audibly present and remains a coherent acoustic source throughout.
AVC Yes — The visible plucking of the harp strings is synchronized with the audible notes; no mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The drummer's arms and legs are visually present and consistently interact with the drum kit throughout the clip.
AES Yes — The drum sounds are audibly present, continuous, and consistent with the visible instrument.
AVC Yes — The visible strikes on the drums and cymbals are synchronized with the audible drum hits.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The train is visually present and maintains a consistent identity and motion throughout the clip.
AES Yes — The train's rolling sound is continuously audible and consistent with the vehicle's movement.
AVC Yes — The rhythmic clacking sounds align well with the continuous forward motion of the train; no mismatch observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The individual is visually present and maintains a consistent appearance and position throughout the clip.
AES Yes — The sound of the harp being played by the individual is consistently audible without unexplained interruptions.
AVC Yes — The individual's hand movements on the strings are synchronized with the audible harp notes.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The engine is clearly visible and maintains a consistent appearance and position throughout the clip.
AES Yes — The engine sound is coherent and consistent until it stops.
AVC Yes — No checkable event; there are no visible actions like operating controls to synchronize with the engine's revving and stopping, and no clear mismatch is observed.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The train, which is the source of the horn, is consistently visible and moves coherently across the scene.
AES Yes — The train horn sound is clearly audible and maintains a consistent timbre across its intermittent blasts.
AVC Yes — No checkable event; the intermittent horn blasts are plausible for the continuously moving train, with no observed mismatch.
input (object absent)
GT (object present)
OURS refwarm ep019
InstructAV2AV (1st frame)
AVI-Edit (1st frame)
VES Yes — The drummer is visually present and consistently playing the drum kit throughout the clip.
AES Yes — The sounds of the drums and cymbals are audibly present and consistent with the action.
AVC Yes — The audible drum and cymbal hits are synchronized with the visible actions of the drummer striking the instruments.