Gan Spaces

»Gan Spaces« is a glance at creating AI-guided visual explorations using a neural network as an origin of inspiration. To construct the synthetic images that formed our source material, we used VQGAN (Vector Quantized Generative Adversarial Network), guided by CLIP (Contrastive Language–Image Pre-training) inputs. The resulting images and animations, resembling a creative erosion, were used as texture components for further use in a 3D environment. It is inspiring to combine machine learning tools with human imagination, guiding each other to create intriguing digital worlds. (Jan. 2022)

Concept, Art Direction

fakenew Studio

Design, Animation, Production

fakenew Studio

Sound

fakenew Studio

Credits

Thanks to the ML community:

Open.ai, Katherine Crowson, Daniel Russel, et al.

»Machine intelligence is the last invention that humanity will ever need to make«

Nick Bostrom, 2015