Summary
Sber has released Kandinsky 6.0 Video through GigaChat, offering text- or image-guided clips with synchronized speech, music and environmental sound at up to Full HD. Sound-enabled generations are currently limited to five seconds, with SD and HD options also available. Sber says the model produces 44 kHz audio, improves real-world motion and beats Kandinsky 5.0 in 71% of comparisons across its evaluation criteria. The company also says the audio component was trained on more than 20 million videos. Developers can use the model under an MIT license, with integrations announced for tools including Diffusers, ComfyUI, FAL and FastVideo. These specifications and comparisons come from Sber and have not been independently validated in the cited coverage.
Hana’s perspective
The significant move is not Full HD. It is the decision to generate picture, speech, effects and music as one synchronized object. For animators and microdrama teams, that compresses several early-production steps into a single prompt: visual blocking, lip-sync, ambience and temp score can emerge together. Five seconds is too short for finished narrative work, and the performance claims remain company claims rather than independent tests. But the MIT license matters. It lets developers test the model inside real pipelines instead of treating it as a closed demonstration. If the integrations arrive as promised, the useful question will be whether creators can preserve characters, timing and editorial control across many shots—not whether one clip looks impressive.
Are you a writer with a great story? We'd love to hear from you: APPLY HERE