Sber Introduces Kandinsky Video 1.1 Beta Version
Sber officially announced the beta rollout of its neural network, Kandinsky Video 1.1. The updated model creates 6-second videos based on textual description or a static image. You can evaluate the algorithm performance on the fusionbrain.ai platform, and in the Kandinsky Telegram bot.
The Features of Kandinsky Video 1.1
The neural network generates a continuous 6-second scene with object and background motion at a speed of 8 or 32 frames per second. The model supports creation of videos in 16:9, 9:16, and 1:1 formats. The updated algorithm has the capability of generating videos using static images, in addition to text descriptions. Thanks to this, users have more options for realizing their creative ideas. Furthermore, users can manipulate video dynamics by changing the “motion score” parameter.
Sber’s Technology Advancement
“We have taken another step into the future of video creation today. Every Kandinsky Video user now has the power to bring their ideas to life in video format. Less than a year since the neural network’s initial launch, our team has significantly improved key parameters such as quality and speed of full-fledged video generation, thereby opening up limitless horizons for creativity.”, said Andrey Belevtsev, Senior Vice-President and Chief Technology Officer of Sberbank, commenting on the new algorithm launch.
