An improved version of the Sber neural network can create 6-second videos on a variety of query topics
The Sberbank team presented Kandinsky Video 1.1 – an updated version of the proprietary neural network for creating full-fledged videos using text descriptions and images. The test version has become available to all users without restrictions.
Format of the generated video — it is a continuous scene with movement in both subject and background. The model generates video sequences lasting up to six seconds at a frequency of 8 and 32 frames per second. With the updated Kandinsky Video, you can create videos in different formats — 16:9, 9:16 or 1:1.
The new version of the model allows you to generate videos not only from text, but also from images, that is, «revive» static picture. In addition, in the new version of the model you can control the dynamics of the generated video using a special parameter motion score.
You can join and try out the capabilities of the Kandinsky Video neural network on the fusionbrain.ai platform and in the official Kandinsky Telegram bot.