ComfyUI is an open-source application for generating images, video, 3D models and audio with AI models, made by Comfy Org. It has a node-based interface: the user builds a workflow as a graph on a canvas, where each node is one step (load a model, encode a text prompt, generate an image, upscale it) and connecting lines carry data from one node to the next.
ComfyUI runs on the user's own computer, on Windows, Linux or macOS with NVIDIA, AMD, Intel or Apple Silicon GPUs. It supports many open models, among them Stable Diffusion, Flux and Wan, plus LoRAs, ControlNets and upscalers. It runs workflows in a queue, re-runs only the changed part of a graph, and saves workflows as JSON. Custom nodes written by third parties extend it, and partner nodes call paid closed models such as Nano Banana and Seedance. A local API lets other programs run workflows.
ComfyUI is free and open source under the GNU General Public License v3.0. Comfy Cloud is a paid hosted version.