Kandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio Generation

Kandinsky 6.0 Video releases open 3B and 29B models for synchronized text- and image-to-audio-video.

Kandinsky 6.0 Video is a family of diffusion models for synchronized audio-video generation: Lite at 3 billion parameters and Pro at 29 billion. Both generate 5-second clips with 44 kHz audio, including lip-sync, in text-to-audio-video and image-to-audio-video modes, and a super-resolution model raises output to 1920×1080. A dual-stream CrossDiT links a pretrained video stream and a newly trained audio stream with bidirectional cross-attention, followed by supervised fine-tuning, reinforcement-learning post-training, and distillation. Side-by-side human evaluation says Pro beats Kandinsky 5.0 Video Pro and is competitive on speech quality; code, checkpoints, and Diffusers integration are released under MIT.