Performative Ai

The similarity in form and function between the façade of high-rise buildings and the structure of cells became the inspiration and initial idea for my design. I wanted to use StyleGAN to fuse the façade of high-rise buildings, which represent artificiality, and the structure of cells, which means nature, to produce images with texture and unique forms. I cut 12 of the most interesting photos into countless square images and arranged them into a dataset. These photos were used to train a StyleGAN model with over 3000 steps in RunwayML. It was fascinating to watch these photos gradually being fused and eroded.

Of the two videos in the final output, I chose the one with the rich texture in order to generate the required depth for the 3D model in Houdini. Firstly, I used RunwayML to create the depth maps. I then generated the required VDB and ABC files in Houdini and Processing. The rendering animation in Unity looks very attractive with its red and blue gradient backgrounds. The 3D models generated by StyleGAN glow in a mysteriously and beautiful way.

--- Growing, Growing, Brighter Everyday ! ---