Meet the team that turns your ideas into work.

Discover EightBots
EIGHTBOTS
Contact salesSign up
Kandinsky LabKandinsky 4

Kandinsky 4.0 T2V Flash

T2V Flash is a trained distillation of the Kandinsky 4 video generator. Its transformer uses an MMDiT-like architecture, paired with a CogVideoX VAE and T5 text encoder.

Distillation changes the generator weights

The developers used latent adversarial diffusion distillation. The release combines a trained generator with a discriminator during distillation; its speed is not just a serving-tier label.ai-forever/kandinsky-4-t2v-flash — model card and files

Credit the upstream components

Practical guidance. CogVideoX and T5 are supporting components. Their inclusion does not make them Kandinsky-developed base models.

Specifications & access

Kandinsky 4.0 T2V Flash specifications
DetailValue and conditions
AccessDownloadable
Model identifiersai-forever/kandinsky-4-t2v-flash
Documented output12 seconds at 480pai-forever/kandinsky-4-t2v-flash — model card and files
Card licenseApache 2.0ai-forever/kandinsky-4-t2v-flash — model card and files