Pixray uses the CLIP models released by OpenAI earlier this year to determine what a named item should look like.

• Perception Engines which uses image augmentation and iteratively optimises images against an ensemble of classifiers
• CLIP guided GAN imagery from Ryan Murdoch and Katherine Crowson as well as modifictions such as CLIPDraw from Kevin Frans
• Useful ways of navigating latent space from Sampling Generative Networks (more to come)