Created by Christian Mio Loclair (Waltz Binaire), ‘Blackberry Winter’ is an investigation into the possibilities of identify motion as a continuous walk in a latent space of situations. The work is a triptychon of artificial human motion in asymmetry. The team developed three different choreographies of human bodies and their ongoing neural relationship in reference to each other, using their custom machine learning solution to weave spatial information into GAN technology.
A custom GAN pipeline called RayGan was developed to teach their network a possibility space of human poses. The ongoing journey through this synthetic space of physical human nature, is then perceived as dance. The training data is generated by 120.000 postures of human bodies, but it was never introduced to a sequence of movements or velocity. All textures, colors and gradients are generated through their designer GAN – a network trained with a curated dataset of visual artists.