Comfy-Org/ComfyUI

ComfyUI is a modular graphical user interface, backend runtime, and local execution engine designed for generative diffusion models. The GitHub snapshot shows the repository is not archived and records a GPL-3.0 license. Designed for visual professionals who demand granular control over generation parameters, the platform provides a modular node graph interface. Users can visually build, execute, and automate multi-stage generation pipelines covering image synthesis, video rendering, audio processing, and three-dimensional model creation without writing code, while retaining local API endpoints for production workflow integration.

The README says some models can run with 4GB VRAM and 8GB RAM using weight streaming; actual requirements depend on model and workflow. Supported hardware includes NVIDIA, AMD, Intel, Apple Silicon, and Ascend GPUs. For installation, regular and new users are advised to use the official desktop application on Windows and macOS. Standalone portable Windows archives exist for NVIDIA, AMD, and Intel GPUs, but the documentation explicitly notes that portable builds are not recommended for regular users. Command-line users can install ComfyUI via comfy-cli by executing pip install comfy-cli followed by comfy install. Alternatively, manual installation uses a git clone, a compatible Python version, suitable PyTorch wheels, and running pip install -r requirements.txt. For another image-generation interface, compare Stable Diffusion web UI; supported models, features, and workflows differ.

Data as of 2026-11-03: 144,726 stars · 17,252 forks.

GitHub{references}