Every major diffusion model, diagrammed and frozen in PNGs

What it does This repository is the official companion cookbook for ComfyUI, collecting reference workflows across image generation, video, audio, and 3D synthesis. Rather than written tutorials, it stores complete node graphs inside ordinary image files. Drag any example PNG into the ComfyUI canvas and the application reconstructs the exact workflow used to generate it.

The interesting bit The README functions as a high-speed index to dozens of techniques—Flux, SDXL, ControlNets, inpainting, model merging, Wan video, even Nvidia Cosmos—with each entry represented by an image that is simultaneously its own source code. It is documentation by artifact.

Key highlights

• Covers a sprawling model zoo: Stable Diffusion variants, Flux, Hunyuan video/audio, LTX-Video, Mochi, AuraFlow, Omnigen2, and more.
• Each example image embeds the full workflow metadata, eliminating manual node reconstruction.
• Includes niche techniques like noisy latent composition, GLIGEN, and area composition.
• Links to third-party workflow hubs and custom node managers for extended use.
• Acts as a living compatibility tracker for what ComfyUI currently supports.

Caveats

• The basic tutorial is explicitly marked “early and not finished.”
• The repository is essentially an index; actual learning requires importing images into the main ComfyUI application.
• Linked documentation is labeled WIP.

Verdict Use this if you already run ComfyUI and want a quick, visual reference for new models or techniques. Look elsewhere if you need standalone tutorials or detailed written explanations; this is a library of puzzles, not prose.