Text to Image Transformers (DALLE)

One of the largest advancements in the generative AI spaces in the last few years was GPT3, which was developed by OpenAI. One of the off shoots of the GPT framework is DALLE, which is able to generate very unique sets of images based on a text prompt. Instead of training strictly based off of text, DALLE is trained off of text to image pairs. 

We leverage a DALLE-like transformer model (not DALLE) in order to create unique styles imbued from the training data. What is more meta than creating NFTs from NFTs?

Each image that is generated is unique in nature and is generated using a random walk. In essence, once an image is generated via the bijutsu.ai art engine, subsequent runs will emit a completely different image.