Kandinsky 6.0 Generates Video and Audio Together
Kandinsky Lab's new foundation models produce synchronized sound and motion from a text prompt, in 3B and 29B sizes.
Kandinsky Lab has introduced Kandinsky 6.0 Video, a family of foundation models built to generate video and audio in one pass rather than stitching a soundtrack onto silent footage after the fact. The project is detailed in a paper hosted on Hugging Face.
The release comes in two sizes: a lightweight 3B variant positioned for accessibility and a larger 29B Pro model aimed at higher-quality output. Both target short-form generation, with clips running up to roughly five seconds and audio produced at a 44kHz sample rate.
Why it matters
Most open text-to-video systems treat sound as a separate problem, which makes lip movement, footsteps, and ambient effects hard to line up with on-screen action. By modeling the two modalities jointly, Kandinsky 6.0 aims for native synchronization — a capability that has largely been the domain of closed commercial systems.
A few things worth noting about the launch:
- Two model tiers (3B Lite and 29B Pro) let teams trade compute for fidelity.
- Synchronized audio-video generation is the core differentiator, not an add-on.
- Short clip lengths place it squarely in the current open-source video generation cohort.
The licensing is listed as custom rather than a standard permissive license, so teams planning production use should read the terms closely before building on it. For researchers, the paper and model family offer a concrete look at how joint audio-video modeling can be approached at open scale.
Sources
- Visit
Kandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio Generation
HF Papers
More from Unknown
All Unknown releases →Liquid AI's LFM2.5-VL-DSpark targets faster VLM inference
The new vision-language model from Liquid AI is tuned for accelerated inference, extending the company's LFM2 line into multimodal territory.
Nari Labs Ships Qwen3-Based TTS and ASR Models
The startup pairs speech synthesis and recognition built on Qwen3, pitching accuracy, low latency, and lower cost.
ZGCM-1 arrives as a fully open 7B reasoning model
A compact foundation model targets math reasoning and agentic search with tool use, and its makers are releasing it fully open.
More in Text → Video
All Text → Video →Apple's FastFrames-L Targets Real-Time Video Generation
The company's new autoregressive text-to-video model aims for interactive latency, a notable shift for a field still dominated by slow batch rendering.
NVIDIA's Cosmos 3 Brings Open Interactive World Models
The new open release pushes beyond passive video generation toward simulated worlds that respond to input.
NVIDIA Releases Cosmos 3 for World Simulation
The new open models push NVIDIA's world-foundation effort toward interactive, real-time video generation.
0 comments
No comments yet. Be the first to weigh in.