Comfy-Org/ComfyUI
A drag-and-connect tool that lets you build image, video, and audio AI pipelines without writing code
ComfyUI is a graphical program that runs generative AI models for images, video, 3D, and audio by connecting nodes on a graph instead of writing code. It natively supports a wide range of open-source models like Stable Diffusion, Flux, Wan, and HunyuanVideo. It runs locally on Windows, Linux, and macOS across Nvidia, AMD, Intel, and Apple Silicon hardware, or in the cloud.
What it does
- What: ComfyUI is a creation engine that lets users assemble image, video, 3D, audio, and text generation models on a node graph (a visual flowchart of connected task boxes) instead of writing code.
- How: Each function, such as loading a model, inpainting, upscaling, or applying a LoRA, is represented as a node that users connect by dragging wires. The system only re-executes parts of the graph that changed since the last run, speeding up iterative work.
- Result: It natively supports current open-source models spanning Stable Diffusion 1.5 through SDXL, Flux.1, Flux.2, Qwen Image, Wan, and HunyuanVideo 1.5, and it is distributed as a desktop app, a Windows portable package, a manual install, or a paid cloud version.
- Extra: Workflows can be saved as JSON files, and a generated image itself can be dragged back in to fully recover the workflow and seed used to make it. Everything runs offline by default, and the `--disable-api-nodes` flag can fully disable the optional paid API nodes.
Why it matters
ComfyUI is one of the most widely used open-source tools for running image and video generation AI, letting people without coding skills combine the latest generative models on their own. With over 129,000 GitHub stars, it is used by individual creators and production pipelines alike.
Terms in this repo
- Node/graph interface · A visual programming style where tasks are shown as boxes (nodes) connected by wires to define a workflow
- LoRA · A lightweight fine-tuning method that adds a small amount of extra training data on top of an existing AI model to change its style or subject
- VAE · A neural network component used to compress images into a smaller representation (latent space) and reconstruct them back
- Inpainting/outpainting · Inpainting erases and refills part of an image; outpainting extends and generates new content beyond an image's original borders
- Quantized model · A model whose internal numbers use lower precision to reduce size and computation cost
Repository description (English)
The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.
Open on GitHubTrending repos
- vorssaint/vorssaint-utilsOne free menu bar app replaces a dozen paid Mac utilities
- Alishahryar1/free-claude-codeA local proxy that lets coding AI agents run on 49 free or cheap model providers instead of one paid service
- freestylefly/awesome-gpt-image-2A library of 532 reverse-engineered prompts that turn GPT-Image2 into a predictable image-making tool
- block/buzzAn open-source workspace where humans and AI agents chat, code, and review in the same rooms
- NousResearch/hermes-agentNous Research's Hermes is an AI agent that gets smarter the more you use it
- virgiliojr94/book-to-skillA tool that turns technical book PDFs into on-demand reference skills for AI coding agents
- VoltAgent/awesome-agent-skillsA single hub collecting over 1000 'how-to' manuals that make AI coding assistants act like experts
- anthropics/claude-plugins-communityA shared shelf where anyone's Claude add-ons get listed for install