Node-based local interface for running image and video diffusion models
ComfyUI is a node-based interface for running image and video diffusion models on your own machine. Instead of a prompt box you wire a graph: load a checkpoint, encode the prompt, sample, decode, then add ControlNet, upscaling or inpainting nodes as needed. Workflows save as files, so a result can be reproduced exactly and shared with others, and a large custom-node ecosystem extends the graph to video, audio and third-party services. It runs locally, which means no per-image fees and no uploads, but also that your GPU sets the ceiling.
Last updated: 2026-09-20. This site only provides an index; for exact features, pricing, and licensing, see the official website.
local image generation, diffusion workflows, controlnet pipelines and reproducible generation graphs
If you're comparing similar products, check the alternatives below, or browse all tools in the AI Models & Platforms category.
It helps a lot. Modern image models want several gigabytes of video memory, and video generation wants much more. Cards with less memory can still run smaller checkpoints and quantised versions, just at reduced resolution or with longer waits.
The application is open source and free to run on your own hardware, and you are responsible for the hardware, electricity and any model licences. A hosted cloud option exists for people who would rather rent a GPU than buy one.
Nodes expose the whole pipeline, so you control sampling, conditioning and post-processing exactly and can reproduce a result later. The trade-off is that nothing is automatic: a prompt-driven tool is faster when you do not care how the image was made.
Developer console and API keys for the Claude model family
Browser playground for prompting and prototyping with Gemini models
Hosted Jupyter notebooks with optional free GPU and TPU runtimes
Where most of the open machine learning world lives