Before The Image

Interactive AI diffusion installation

Before The Image

Real-time Interactive AI diffusion installation

Before the image is an interactive installation where a custom (Low-Rank Adaptation) LoRA model, trained on images of faces in noise, exists in a constant state of metamorphosis, generates visuals in real time through StreamDiffusion.

Visitors can move and interact with the interface to influence the flow of images as they continuously shift between the familiar and the unfamiliar.

Rather than aiming for likeness or realism, the system searches for what exists before the image, the fleeting moment when noise could become anything. Drawing on Sergei Eisenstein’s idea of animation as the plasmatic, matter in continuous transformation, the piece explores AI diffusion model not as a technology for reproducing reality, but as a medium for creative resistance.