Below are select output images from model-1. All inputs, text prompts and outputs were chosen by the artist whose work it was trained on. The code for the model as well as more information on it's training and output can be found in it's documentation.