Sber introduced Kandinsky 2.1, an updated neural network for generating images on demand

Sber announced a major update of the neural network Kandinsky, which generates pictures from a text description.

The neural network can change existing images on request, merge several images, draw fragments, create images in the same style that the user suggested, and more. She understands requests in 101 languages, including Russian and English.

In version 2.1, the neural network received additional training (on 170 million text-image pairs and a separate selection of 2 million pictures). At the same time, we improved the encoding model, thanks to which neural networks began to better give high-resolution pictures, as well as faces and other complex objects.

The neural network is available at site and form Telegram bot, it can also be activated on Sberbank smart devices with the Start Artist command.