Images and video
ComfyUI
A node-based studio for making AI images and video on your own GPU, where you see and control every step.
- Rank
- #4
- Stars
- 145k
- Licence
- GPL-3.0
- Checked
- 6 Nov 2026
What it does
ComfyUI runs image and video models such as Stable Diffusion and Flux locally. Instead of one prompt box, you connect boxes (nodes) for the model, the prompt, the sampler and the output into a workflow you can save and share. That makes results repeatable and lets you chain steps like upscaling or inpainting. Nvidia even connected its RTX Remix modding tool to it.
How to use it
Shell: Windows 11 (File Explorer), Nvidia GPU recommended
Download the Windows portable build or ComfyUI Desktop from the repo's Releases page and extract it.
Put a model file (.safetensors) into the models\checkpoints folder.
ComfyUI_windows_portable\ComfyUI\models\checkpoints\Start it, then open http://127.0.0.1:8188 in your browser and press Run on the default workflow.
.\run_nvidia_gpu.bat
Best practices
- Install custom nodes only from authors you trust. Custom nodes are code, and a malicious one has stolen users' passwords before.
- Use .safetensors model files, not .ckpt or .pt, which can run hidden code when loaded.
- Save the workflow JSON with your images so you can recreate any result later.
- Keep it on 127.0.0.1. Don't add --listen to expose it on your network unless you know who can reach it.
In the news
- Tom's Hardware · 4 Jul 2024 · Nvidia's RTX Remix goes open source, with a REST API that interfaces with ComfyUI
- 404 Media · 11 Jun 2024 · Hackers Target AI Users With Malicious Stable Diffusion Tool on GitHub to Protest 'Art Theft'