//
The most powerful image generation interface
ComfyUI is a node-based interface for AI image generation. It's more complex than Automatic1111 but offers unmatched flexibility and performance. This guide covers everything from installation to advanced workflows.
Clone the repository and install dependencies.
# Windows (with Python 3.10+):
git clone https://github.com/comfyanonymous/ComfyUI
cd ComfyUI
pip install -r requirements.txt
# Or use the portable version:
# Download from ComfyUI releases on GitHubGet base models for your use case.
Models folder structure:
ComfyUI/
models/
checkpoints/ <- Main models (SD, SDXL)
clip/ <- Text encoders
vae/ <- VAE models
loras/ <- LoRA fine-tunes
controlnet/ <- ControlNet modelsStart the server and open in browser.
python main.py
# Opens at: http://127.0.0.1:8188
# For low VRAM:
python main.py --lowvram
# For fp16:
python main.py --force-fp16Start with built-in workflows or download custom ones.
Built-in: Load Default button
Custom: Drag & drop .json or .png with workflow
Popular workflow sources:
• ComfyUI_examples on GitHub
• CivitAI workflows
• OpenArt.ai workflowsLearn the basic node types.
Essential nodes:
• Load Checkpoint - Loads your model
• CLIP Text Encode - Processes prompts
• KSampler - The main generation node
• VAE Decode - Converts latent to image
• Save Image - Saves your output❓ CUDA out of memory
✅ Launch with --lowvram or --novram flags. Reduce batch size. Close other GPU applications.
❓ Missing custom nodes
✅ Install ComfyUI Manager for easy node installation. Or manually clone to custom_nodes folder.
❓ Workflow won't load
✅ Missing nodes cause this. Install required custom nodes or use workflows matching your setup.