Kandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio Generation
Kandinsky 6.0 Video is being released as an open-source audio-video generation model with code, checkpoints, and diffusers support.
The report describes two diffusion models: a 3B-parameter Lite version and a 29B-parameter Pro version. Both generate five-second clips with synchronized 44 kHz audio, including lip-sync, from text or image prompts. A built-in super-resolution model can raise outputs to Full HD. The authors say the Pro model beats Kandinsky 5.0 Video Pro in human comparisons and is competitive with leading audio-video systems, especially on speech quality. HF Daily Papers' note
The report describes two diffusion models: a 3B-parameter Lite version and a 29B-parameter Pro version. Both generate five-second clips with synchronized 44 kHz audio, including lip-sync, from text or image prompts. A built-in super-resolution model can raise outputs to Full HD. The authors say the Pro model beats Kandinsky 5.0 Video Pro in human comparisons and is competitive with leading audio-video systems, especially on speech quality. HF Daily Papers' note
score 7