Sber releases Kandinsky 6.0 Video neural network with synchronized sound
Sber has introduced the Kandinsky 6.0 Video neural network, which is capable of creating video clips with synchronized speech, music, and ambient sounds.
As the developers reported, the new model allows generating clips up to five seconds long in SD, HD, and Full HD (1920 × 1080 pixels) quality. The audio is generated at a sampling rate of 44 kHz, with characters' lip movements synchronized with speech. If desired, a clip can be created without an audio track.
A prompt is formed using a text description or based on an uploaded first frame. The lineup includes two versions: Lite with 3 billion parameters and Pro with 30 billion parameters. According to the company, the model outperformed the previous version in an average of 71% of cases across all criteria of motion and physics reproduction.
The tool is integrated into the GigaChat service and is also uploaded to Hugging Face. The model's source code and weights are open under the permissive MIT license, enabling developers to use it in their own products for free.
What it means
The model is available to users in GigaChat, and third-party developers can integrate it for free under the MIT license via the Hugging Face repository.
Sources
- Hi-Tech Mail: Сбер представил Kandinsky 6.0 Video: генерирует видео с синхронным звуком в Full HD media outlet
- iXBT.com: «ГигаЧат» Сбера научился создавать видео с синхронным звуком media outlet
- 3DNews: «Сбер» выпустил Kandinsky 6.0 Video — ИИ-модель для генерации видео с синхронным звуком media outlet
- Хабр: «Сбер» выпустил модель Kandinsky 6.0 Video для генерации видео с речью и музыкой media outlet
This article was prepared with the help of AI and translated from the Russian original. Spotted an error — let us know.
Published 7 October 2026, 14:32 · No updates · Russian original




Comments on this story live under its post in our Telegram channel. Comment on Telegram