← Back to index

AbstractbyAI

This Machine Learning art turned into freelance work and teaching for great clients. Tools used to create those images are StyleGAN, Google CoLab, Google Drive, After effects, RunwayML, Photoshop.

Very first and still favourite outcome exploring StyleGAN.

01 Creating new shapes and colors

For the first step, I used StyleGAN with a trist. Instead of a big and similar data used usually set I went for a small, diverse set of images and trained it for a fairly short time. This opened doors to new variations in color and form.

Training progress of 1 out of 3 datasets.

02 Feeding the trained dataset back

Next I used the results of my earlier experiments in order to feed them back and train a new model with a more advanced technology using RunwayML.

03 AbstractbyAI by AbstractbyAI: Text to Image

Finally, I utilized a text-to-image tool to generate images in the style of my newly trained dataset. This enabled me to create for example, a "Shanghai Skyline" image in the unique AbstractbyAI aesthetic.

An AbstractbyAI interpretation of "A rapeseed field in Malmö".
"A sunset over Malmö".

Freelance work examples

Finally, I used Stable Diffusion, a text-to-image AI tool, to generate images in the style of my previous dataset. This enabled me to create for example a "Shanghai Skyline" image in the unique AbstractbyAI aesthetic.