What it is
ComfyUI lets you generate images on your own computer with open image models. Instead of a single prompt box, you build a workflow of connected boxes (nodes): load a model, write the prompt, set the size, generate, upscale... It looks complex at first, but ready-made templates make it easy to start, and it gives far more control than most online generators.
What you need
- A dedicated NVIDIA GPU with 8 GB+ of VRAM is ideal. Apple Silicon Macs work; AMD works on Linux (and increasingly on Windows). CPU-only is possible but very slow.
- 10–30 GB of free disk space for models.
Install
Easiest: download ComfyUI Desktop for Windows or macOS from comfy.org and follow the installer.
Manual install (all systems, needs Python and git):
git clone https://github.com/comfyanonymous/ComfyUI
cd ComfyUI
pip install -r requirements.txt
python main.pyThen open http://127.0.0.1:8188. (NVIDIA users should first install the CUDA version of PyTorch, as described in the README.)
Use it
- Open the Templates (workflow browser) and choose a basic text-to-image template.
- If the template needs a model you don't have, ComfyUI offers to download it. Manually downloaded models go in
ComfyUI/models/checkpoints/. - Type your prompt in the positive prompt box (and what to avoid in the negative one).
- Click Run / Queue. The result appears in the Save Image node and in
ComfyUI/output/.
a red fox sitting in fresh snow at sunrise, soft golden light, detailed fur, photo, 85mm lensTip
Drag a ComfyUI-generated PNG back into the window: the full workflow that created it is stored inside the image.
Limitations
- Steep learning curve for advanced workflows.
- Image quality and speed depend heavily on your GPU.
- Check each model's licence before commercial use.
Comments 0
Share tips, prompts that worked for you, or questions.
No comments yet. Be the first to share a tip.
Log in or create a free account to comment.