A person taps a door gently with one knuckle, producing a soft knock. Then pounds the same door with a fist, producing a loud deep boom.
Seedance 2.0ByteDance · proprietary
Kling 3.0 OmniKuaishou · proprietary
Veo 3.1Google · proprietary
LTX-2.3Lightricks · open
Ovi 1.1character.ai · open
JavisDiT++JavisDiT · open
MagiHumanGAIR-NLP · open
Per-prompt rubric
Basic standards
Visual presence
- a person
- a door
Event: a person strikes a door with a hand
Audio presence
- a door
Sound: a hand impact on a door
Key standards
Visual physical commonsense (V-PC)
- The person strikes the door twice with a hand, first as a gentle knuckle tap and then as a forceful fist pound, with each hand impact visibly contacting the door itself.
Audio physical commonsense (A-PC)
- The forceful fist pound is noticeably louder than the gentle knuckle tap.
- Each impact sound is a real hand-on-door impact.
Cross-modal physical commonsense (AV-PC)
- Each impact sound aligns in time with the visible contact of the hand on the door.
- The louder impact corresponds to the forceful fist pound.
- The quieter impact corresponds to the gentle knuckle tap.