A person speaks normally, then inhales SF6 gas and speaks again. Their voice drops dramatically to a deep gravelly tone.
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
- an SF6 gas source
Event: a person inhales gas and speaks
Audio presence
- a person
Sound: human speech
Key standards
Visual physical commonsense (V-PC)
- The person performs a visible inhalation action from a held source before the second speaking segment.
- The held source is a plausible gas-delivery container with no liquid contents.
Audio physical commonsense (A-PC)
- The post-inhale voice is noticeably deeper and more gravelly in timbre than the pre-inhale voice.
Cross-modal physical commonsense (AV-PC)
- Each speech segment in the audio coincides in time with the person's visible mouth movements.