Kandinsky 4.0 Video
The project describes text and image-conditioned video generation. Its public code repository releases Flash and V2A, while the full T2V and I2V entries remain marked as forthcoming.
Access not confirmed
The full generator is distinct from Flash
The research page describes a diffusion transformer, a CogVideoX-derived video VAE and T5-V1.1-XXL text embeddings. Its original model description includes twelve-second HD video.Kandinsky 4: research project and architecture
Historical product links do not establish a current version selector
The project page links a hosted generator and describes an API. A current route that selects these exact full-model weights was not established in this review.Kandinsky 4: research project and architecture
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Access not confirmed |
| Model identifiers | Kandinsky 4.0 T2VKandinsky 4.0 I2V |
Related models
- Kandinsky 4.0 T2V Flash
Kandinsky Lab ยท Kandinsky 4
A distilled Kandinsky 4 model for twelve-second 480p video, with a CogVideoX VAE and T5 encoder.
Launch date not confirmedDownloadable