Sber’s neural network users created more than 200 million images in 2023

No time to read?
Get a summary

According to the report of the bank’s press service, more than 12 million users used Sber’s Kandinsky neural network in 2023. They have created over 200 million images based on text queries.

“The model ranked first in terms of growth rate and second only to Stable Diffusion in terms of popularity among developers, according to Hugging Face, an AI source of the best open source solutions,” the message says.

As Sberbank explains, Kandinsky understands requests from a wide list of topics in more than 100 languages. Using a neural network, users can create an unlimited number of photorealistic images using different styles.

The neural network can also replace individual objects and entire areas in the picture, blend several images, complete the image, and create images in infinite canvas mode. The model also knows the local cultural code. He is familiar with architectural landmarks, objects and elements of folk art.

Users also have the ability to create four-second animated videos and full-length videos up to eight seconds long using a neural network.

Let us remind you that Kandinsky works in the GigaChat service, using the SberBank Online and Salute mobile applications, as well as on Sber smart devices using the voice command “Start artist”. The model was developed and trained by Sber AI researchers with partnership support from scientists from the AIRI Artificial Intelligence Institute on the combined dataset of Sber AI and the company SberDevices.

No time to read?
Get a summary
Previous Article

Russian army suppresses GPS signal in half of Poland

Next Article

David Rusell, Alicante Classical Guitar Award