This project is heavily based on "Dino Diffusion": Bare-bones Diffusion Models by Ollin Boer Bohan.

• Uses p5.js v.1.10.0 by the Processing Foundation
• Uses the ONNX Runtime Web v1.18.0 by Microsoft

To understand how diffusion models generate images, Ollin Boer Bohan wrote PyTorch code to train an extremely simple, bare-bones diffusion model. Bohan trained a tiny diffusion model that generates 512×512 botanical images in the browser. You can play with Bohan's demo here. The version in this project is a port of Bohan's JavaScript-based front end to p5.js, allowing for new types of tinkering.

A p5.js demo of this project can be found here, at OpenProcessing.org. Note that this demo loads a trained model file which is approximately 8MB.