This work studies different metamorphosis directions of the same starting point in the latent space of a trained GAN model. Six parallel timelines are displayed from the infinite set of direction vectors, all of them exploring a different range of possibilities.

The subject of the study is a seemingly organic ambiguous structure that has emerged like a weed during training of the GAN model on a set of film photographs representing humans and man-made structures.

AI motion painting. Custom GAN trained on photography collection. AI-generated sound. 2022 Ivona Tau