# ComfyUI

A node-based visual editor for building and running diffusion-model workflows that generate images, video, 3D models and audio, usable through a desktop app, a manual install or a hosted cloud version.

ComfyUI is a visual, node-graph interface for diffusion model workflows. Users connect nodes to build pipelines that generate images, video, 3D models, audio and text without writing code, and can reuse workflows as subgraphs or templates.

It supports current open-source diffusion models locally, including on low-end hardware such as 4GB of VRAM with 8GB of RAM through asynchronous weight streaming, and partner nodes add access to additional closed-source models. Finished workflows can be exposed as a simplified UI through App Mode, or called through API endpoints for use in other applications.

It runs on Windows, Linux and macOS as a desktop application or manual install, and is also available as a paid cloud version. It is written in Python and licensed under GPL-3.0.

Website: https://www.comfy.org/