ComfyUI is an open-source, node-based engine for generating images, video, audio and 3D with AI models. Instead of a single prompt box, you build a graph: nodes that load a model, encode a prompt, sample, decode and save, wired together on a canvas (ComfyUI README).
What it runs
The README lists native support for image models such as Stable Diffusion 1.5, SDXL and Flux, video models such as Wan 2.1 and 2.2 and LTX-Video, plus audio and 3D models. It runs on Windows, Linux and macOS, as a desktop app, a Windows portable build, or a manual install, and it supports NVIDIA, AMD, Intel and Apple Silicon hardware.
What a workflow is
The default workflow is a handful of nodes: a checkpoint loader, two text encoders (positive and negative prompt), an empty latent image, a KSampler, a VAE decode and a Save Image node. Workflows save as JSON, and a generated PNG carries its full workflow, so dragging the image back onto the page restores the graph and seeds.
ComfyUI is released under the GNU GPL v3, and custom nodes (third-party Python packages) extend it. The ComfyUI custom node install guide covers the eight packs most setups add.
Next steps
- How to use ComfyUI walks through a first generation.
- ComfyUI system requirements lists what the machine needs.
If you would rather not set the machine up yourself, the ComfyUI template on Aquanode provisions the box, installs ComfyUI and starts it when you deploy it; once the deployment is Ready, the Overview tab shows the ComfyUI URL.
Sources
- ComfyUI README. github.com
- ComfyUI LICENSE. github.com
- ComfyUI basic_api_example.py (default workflow nodes). github.com