Based and trained on artworks by Tatiana Plakhova (Complexity Graphics)

Model training process:
step by step

The model was gradually trained on artworks by Tatiana Plakhova (Complexity Graphics) during 1 week, showing with every iteration more steady images (detalisation, composition, colors). Eventually the results reached the level of Tatiana’s works: the neural network successfully expressed distinctive and unnoticeable features of her individual style (which has been forming for more than 15 years!). Neural networks don’t reach the ability to exactly replicate an artwork of the ‘teacher', nevertheless they consistently develop new distinctive features

01 iteration
02 iteration
03 iteration
04 iteration
05 iteration
06 iteration
07 iteration
Three things can endlessly keep attracting our attention: fire, water and metamorphosis created with StyleGAN3.
Artists have a powerful control over results by choosing input images and their place in the composition. Such a flow of transformations can open new opportunities for inspiration and artistic exploration.
Influence of parameters
Visuals depend on the value of parameters. Little by little work with artificial intelligence gets predictable and manageable, but still doesn’t lose its miracle — machine’s imagination develops further. Here you can see how far you can go from the dataset and first results.
We experiment with final ‘artificial' results by synthesizing them with different forms and putting them into unimaginable context.

Share experiment


Denis Leonov, Max Kan, Vadim Kondrattsev, Ksenia Telegina
Design & Experiment lab:

Katya Janzen, Kamila Gizitdinova
Art Direction:

Oleg Yusupov, Tatiana Plakhova
Workflow in Phygital+