Note This project is an open interactive art canvas

Note This demo was useful to convert SD 1.x prompts syntax to SD 2.x syntax

Note Same as v.1 but compatible with Stable Diffusion 2

Note Big milestone! Dreambooth finally allowed to easily train custom styles and concepts, fast and cheap.

Note Outpainting !

Note CLIP Interrogator is the reverted process, and let you retrieve the prompt that would have been use to generate the same kind of image than your input image

Note This text-to-Pokémon demo was the first to show that we could train specific dataset of images and had a huge impact.

Note This at the time, was the first doodle app to ever let you draw a very simple image, and diffuse an entire image according to your input drawing

Note A game, based on Wordle, where you had to guess what was the prompt used to generate a 9 images grid with DALL-E

Note The GOAT

Note The Diffusers Gallery is a library of fine-tuned models on custom and well-know style, trained with Dreambooth, and later LoRa