ComfyUI
A node graph for image and video models. Steeper to learn than anything else here, and the only one that lets you build exactly the pipeline you want.
- Self-hosted
- Active
- Allowed adult content
- Open source
At a glance
- Where it runs
- Self-hostedYou run it on your own machine. Nothing leaves it unless you send it.
- Adult content
- AllowedAdult content is permitted within the published rules.
- Trains on your content
- NoYour content is not used to train models.
- Status
- ActiveMaintained and working today.
- Account needed
- No
- Licence
- GPL-3.0
- Cost
- Free and open source. A paid hosted version exists separately.
- What it keeps
- Nothing leaves your machine. Images and workflows are written to local folders.
What it takes to run
- Graphics memory
- 4 GB VRAM, 12 GB recommended
- System memory
- 8 GB, 16 GB recommended
- Disk
- about 20 GB
- Runs on
- Linux, macOS, Windows
Find what fits your machine on the tools page — pick your hardware from the first menu.
You connect boxes together: a model loader into a sampler into a decoder, with whatever you like in between. The result is that anything the underlying libraries can do, ComfyUI can be made to do — and that new model architectures usually get a ComfyUI workflow before they get a button anywhere else.
The cost is the learning curve. Your first hour is spent discovering that a workflow is a file, that other people's workflows arrive embedded in the PNGs they generated, and that the custom node manager is not optional.
It runs on your machine and imposes no content policy of its own. What the model you loaded will do is between you and the model.
Worth knowing before you start
- A real learning curve The interface assumes you know what a sampler, a VAE and a CLIP encoder are before you start.
- Custom nodes are third-party code Installing one runs somebody else's Python on your machine with your permissions. source
What it does
- Node graph workflows
- Workflows embedded in output PNGs
- Custom node ecosystem
- Queue and batch
- API for scripting
Models it runs
How it relates to other tools
- required by Krita AI Diffusion — installs and drives its own copy if you do not have one
- required by AnimateDiff — in practice, most people drive it from ComfyUI
- required by HunyuanVideo — the usual way to run it on a desktop
Signs of life
Checked automatically. These are the only figures on this page a machine wrote, and they say when they were taken.
- Last answered
- Yes, 3 hours ago Its website responded when we asked.
- Stars on GitHub
- 133,408 15,782 forks.
- Last commit
- 4 hours ago2026-09-16 01:43 UTC
- Latest release
- v0.36.0
- Archived copy
- Wayback Machine A snapshot, in case this one stops answering.
Categories
Something here wrong or out of date? Tell us — this page is only worth having if it is right.