ComfyUI
The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
What is ComfyUI?
The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Overview of ComfyUI
What it is
ComfyUI is an AI creation engine for visual work, described in its README as a modular diffusion model GUI, API and backend built around a graph/nodes interface. It is available on Windows, Linux, and macOS, either locally through a desktop application or a portable install, or on Comfy Cloud. Manual installation supports all operating systems and GPU types, including NVIDIA, AMD, Intel, Apple Silicon, and Ascend. The project is developed across three repositories: ComfyUI Core, Comfy Desktop, and ComfyUI Frontend, with a weekly release cycle targeting Monday that the README notes changes regularly.
Key features
- A visual node graph for building and reusing image, video, audio, 3D, and text workflows without code.
- Reusable subgraphs, workflow templates, App Mode, and a local API for integrating workflows into applications.
- Local execution with asynchronous queueing, partial graph re-execution, VRAM and RAM management, model offloading, and support for quantized models.
- Native model support spanning image generation and editing, video generation, audio generation, 3D and vision, and text generation, with a workflow library of ready-to-run templates.
- Built-in tools for inpainting, outpainting, reference conditioning, masks and compositing, model merging, upscaling, frame interpolation, segmentation, depth estimation, and media processing.
- Save and load workflows as JSON, or recover complete workflows and seeds from supported generated media; high bit depth formats such as 16 bit PNG, 32 bit EXR, and 10 bit AVIF are supported, along with HDR video and images.
- Runs fully offline by default;
--offlinedisables the optional paid Comfy API nodes, and--disable-partner-nodesdisables only the API nodes. Custom nodes extend functionality, andextra_model_paths.yamlconfigures additional model locations.
Use cases
- Generating images, videos, 3D models, audio, and text with open-source models.
- Exposing sophisticated workflows through a simple UI using App Mode.
- Integrating workflows into production pipelines through API endpoints.
- Running large open-source models on consumer hardware; the README states models can run on as low as 4GB VRAM plus 8GB RAM using asynchronous weight streaming.
- Reaching closed-source models such as Nano Banana, Seedance, and Hunyuan3D through partner nodes.
Getting started
The README does not include install or quick-start commands. It points to three routes: the desktop application for Windows and macOS, a manual install supporting all operating systems and GPU types, and Comfy Cloud. See the repository README for the current instructions.
License and pricing
ComfyUI is released under GPL-3.0 and is open source. Comfy Cloud is described as the official paid cloud version for users who cannot afford local hardware; the README does not state pricing.
Based on the project's public README and repository metadata.