MetalGAN
generated visuals, 2019
Experimental workflow of black metal visuals production using GANs in RunwayML. Detailed description of the process here︎︎︎.
Featured in https://runwayml.com/madewith/.



The final aesthetic is achieved by chaining 3 models in Runway: DeepLab (which extracts semantic map of the nature from a source image), SPADE (which generates synthetic nature from the semantic map) and Arbitrary-Image-Stylization (which transfer the aesthetic style from one image to the sequence of synthetic nature).
Different style transfer:

