ComfyUI — Getting Started

• Windows 10/11; NVIDIA GPU recommended.
• Python 3.10 or 3.11 (for manual install).
• Disk space for models (several GB+).

1. Download the ComfyUI Windows portable release (.zip).
2. Extract to a simple path (e.g., D:\ComfyUI).
3. GPU: run run_nvidia_gpu.bat. CPU/AMD: run run_cpu.bat.
4. Open http://127.0.0.1:8188 in the browser.

PowerShell in your target folder:

# Create and activate venv (Python 3.10 or 3.11)
py -3.10 -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip

# Get ComfyUI
git clone https://github.com/comfyanonymous/ComfyUI
cd ComfyUI

# Install dependencies
pip install -r requirements.txt

# (GPU) Install PyTorch with CUDA that matches your driver
# Get the correct command from pytorch.org (example for CUDA 12.1):
# pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

# Run
python main.py

Put files in these folders inside ComfyUI:

• models/checkpoints — base models (.safetensors, .ckpt)
• models/vae — VAEs
• models/clip, models/unet — as required

• Stability Matrix: curated tools, builds, and guides — stabilitymatrix.ai
• CivitAI: models, LoRAs, embeddings — civitai.com
• Hugging Face: model repositories — huggingface.co
• ComfyUI Wiki — GitHub Wiki
• ComfyUI Manager (community nodes manager) — ComfyUI-Manager

• Portable: run the provided .bat.
• Venv: python main.py
• Open http://127.0.0.1:8188 in the browser.

• OOM/black images: reduce resolution/batch; ensure GPU PyTorch.
• Missing models: verify files in models/* folders.
• Port busy: change port with --port 8189.

# Change port
python main.py --port 8189