Everything is already an image (2022)

In a collaboration with Fieldstation (Faculteit Architectuur), I created a live data visualization that takes images as input and transforms them into another image using several AI transformations. I was using GPT3 or text transforms and stable diffusion for images as well as found data from a public API for calculating latent space vectors of images.

Integrated in the expo of its input images the robot endlessly creates new images from that very same expo

Showing latent space crawling in both image and text dimensions, to explain how an AI “thinks”.