Sber’s Kandinsky neural network has been updated to a new version
[ad_1]
The press service of the credit institution reported that an improved image generation model will be available to users in the near future.
It is noted that the neural network creates images based on text descriptions in Russian and English. In the new version Kandinsky 3.1, the quality of generations has increased significantly. Some artists, designers, and bloggers have already received initial access to the update.
“Today is exactly one year since Kandinsky 2.1 was released. During this time, we have constantly developed our neural network, which helps people create new images and gives absolutely everyone phenomenal opportunities for creativity. Compared to the previous model, Kandinsky 3.1 has become even faster, more convenient and more realistic. Kandinsky 3.1 is a flexible, feature-rich and completely free tool that will turn anyone into an artist and creator. “Soon everyone will be able to test the new capabilities of the neural network,” said First Deputy Chairman of the Board of Sberbank Alexander Vedyakhin. He noted that the updated version will remain free and will be available on various platforms.
The company clarified that the key feature of the update was the increased speed of image generation. The time to create one image has been reduced by almost 10 times, in addition, it has become possible to increase the resolution to 4K. Users will also be able to rediscover the functions of creating different variations of images, mixing pictures and text, creating sticker packs, and the ability to make local changes to the picture without changing the composition of the scene.
Among other things, the team will soon present a new model of the Kandinsky Video 1.1 neural network for generating videos from text descriptions. As experts said, it has significantly increased the quality of generations due to an increase in the volume of the training dataset of text-video pairs and architectural improvements to the model. The updates made it possible to double the video resolution compared to the previous version of the neural network.
Photo: Sber press service
[ad_2]
Source link
