Pre-baked ComfyUI notebooks for the GPU-poor

What it does

This repository is a catalog of Jupyter notebooks for Google Colab that install and launch ComfyUI—the node-based Stable Diffusion interface—already loaded with specific models, ControlNet weights, and community extensions. You select a notebook matching your desired model and stability tier, and Colab provides the GPU.

The interesting bit

The maintainer has pre-solved the dependency puzzle for many popular fine-tuned models (Waifu Diffusion, Arcane style, Midjourney-v4, etc.) and offers three freshness tiers—lite, stable, and nightly—so you can pick frozen reliability or a bleeding-edge daily build. It is essentially a remote GPU rental agency with the paperwork already filled out.

Key highlights

• Three stability tiers: lite (minimal stable setup), stable (adds ControlNet), and nightly (latest ComfyUI, PyTorch 2.0, and daily extension updates)
• Pre-configured notebooks for specific artistic models (e.g., Mo-Di Diffusion, Cyberpunk Anime, PaperCut) with prompt tokens documented in the README
• Includes JSON workflow templates and a YouTube tutorial video
• Runs via Google Colab, so the actual GPU compute is handled remotely

Caveats

• Explicitly marked as 🚦 WIP 🚦, so notebooks and extension versions may shift
• The nightly track uses daily extension updates and the latest ComfyUI, which implies higher volatility than the lite or stable tracks
• The README is mostly a grid of badges and model links; documentation beyond the tutorial video is thin

Verdict

Ideal for Stable Diffusion experimenters who want to try ComfyUI and niche fine-tuned models without a local GPU. Not for developers who already have a working CUDA environment and prefer to manage their own dependencies.