Neon, pixels and Russian cultural code - how "Shedevrum" changes user photos with the help of the YandexART neural network

Now the application allows you to edit your images and photos in eight modes

The Russian mobile application "Shedevrum" has integrated the YandexART neural network into its functionality. Now users can edit their photos and pictures with YandexART thanks to "filterums" - creative modes.

Male portrait from the "Brush" filterum based on a regular photograph

To use filterums, you need to select the one you like and upload a picture or photo from your mobile device to the application. As a result, the image may look like plush or knitted toys, cartoon frames, pixel graphics, paintings drawn with a brush, winter landscapes, neon canvases and colorful meadows.

This is what the "Winter" filterum looks like on a summer landscape

The processing result can be downloaded, made into author's sticker packs, avatars in social networks - or simply published in the "Shedevrum" feed.

Filterum "Plush" - works even on images of animals

YandexART - the working tool of all these creative modes - is a diffusion neural network that creates images and animation in response to text requests.

It [YandexART] was trained on the example of 330 million pictures with descriptions, knows the Russian cultural code and uses a special text recognition algorithm to more accurately understand the wishes of users. YandexART can qualitatively work out small details, work in a given artistic style and create photorealistic portraits. The neural network is integrated into "Shedevrum", used in Yandex Business, and will soon appear in the Keyboard and other services of the company.
Yandex Press Service

Filterums are not the only creative modes of "Shedevrum". Users have already been promised the imminent possibility of creating their own image processing modes using text requests and additions.

Earlier it became known that "Shedevrum" entered the top 3 of AI applications used in the world at the end of 2023. The rating of applications with generative artificial intelligence was published by Data.ai (ex-App Annie).