Sber announced that neural network Kandinsky 2.2 has the ability to create videos based on text descriptions. It is noted that a 4-second video is generated with an animation effect in just 20 seconds.

To generate a video, you need to compile a detailed prompt. Then, the neural network will propose 16 versions of scene animations, and after the user makes their choice, it will generate the video. The video's technical parameters are 24 frames per second and a clearness of 640 × 640 pixels. The service operates in testing mode.

The company promises that by the end of 2023, all users will be able to use this new feature. And now, to get access, you need to leave an application through the Telegram bot. At the same time, it will take some time to process it.

It was previously reported that the Russian company Yandex has introduced a new feature to the YandexGPT neural network - a summary of video clips in Russian.

Sources :
Kandinsky

Сейчас на главной