I am studying how neural networks can be used in character and environment creation. The process involves generating expressive silhouettes and filling them with generated "parts". The methodology is similar to working with photographic sources in concept art — photobashing with additional drawing, processing, and color correction. Of course, the neural network does not understand the language of form, anatomy, and perspective, but it can quickly visualize unusual combinations proposed to it. In addition, interesting ideas can be stumbled upon, interpreted, and quickly sculpted in 3D.