Underwater Abstractions

2025
Moving image [00:38]
2160 x 3840 px

Underwater Abstractions imagines a synthetic ocean in which marine forms never settle into fixed identities. Generated from marine-life image classes using a generative adversarial network (GAN), the moving image passes through shifting configurations of bodies, textures and underwater structures.

Further visual processing draws these forms into a liquid abstraction, softening their contours and allowing colours and surfaces to flow into one another. The result is an ecosystem beyond taxonomy, where traces of recognisable organisms emerge only to dissolve into the next transformation.