Open source · ★ 137,381 GitHub stars (as of 2026-10-14)
ComfyUI
ComfyUI is a GUI and engine for diffusion models that generates images, video, 3D and audio by connecting nodes.
Overview
ComfyUI is a generation engine that creates images, video, 3D models and audio by connecting models and processing steps on a node graph screen. It provides a GUI, an API and a backend for working with diffusion models.
As you connect nodes on an infinite canvas, every model, parameter and processing step is visible on screen. It focuses on running on local hardware, and by loading weights asynchronously in parts, it can run large open models even in environments with little VRAM. Closed models can be used through Partner nodes, and App Mode lets you wrap a complex workflow in a simple interface, or you can attach it to a production pipeline through the API.
It suits visual practitioners such as designers and video creators who want to control every step of the generated output themselves. There is also Comfy Agent, which builds and runs a workflow when you describe the result you want.
What it is used for
- Connect nodes to build image and video generation pipelines.
- Load a community template workflow and generate right away.
- Deploy a finished workflow as an API and call it from your app.
- Turn a complex workflow into a simple interface with App Mode and share it.
Pricing and license
Desktop app is a free download; Comfy Cloud is paid (free trial available). For the license, see the official website
Getting started
- Download the desktop app for Windows or macOS, or sign up for Comfy Cloud.
- Load a community template or create a new workflow.
- Run it, check the results, and adjust nodes and parameters.
Links
Related terms and companies
Sources
Compiled from public sources (official pages, READMEs). Pricing and features change, so check the official site.