How to Install ComfyUI on Windows

ℹ️ Note about updates: For the most up-to-date Desktop installation info, check the official docs.

If you are new to ComfyUI, we recommend using the Comfy Desktop version — it handles everything for you. → Comfy Desktop Installation Guide

If you prefer maximum control or already have a ComfyUI setup, this guide covers:

Official portable version (recommended for most users)
Manual installation (via Git)
Third-party launcher (Aaaki ComfyUI Launcher)

System Requirements for Windows

Component	Minimum Requirement
GPU	NVIDIA GPU with CUDA support, 4GB+ VRAM (RTX 3060 or higher recommended). GPUs with less than 3GB can use the `--lowvram` option.
CPU	Any modern x86-64 processor; CPU-only mode available via `--cpu`
RAM	8 GB minimum, 16 GB recommended
Storage	SSD strongly recommended, at least 20 GB free space (bare minimum for the app — models for common workflows quickly exceed 50–100GB)
Python	3.12 or 3.13 (the portable version includes Python 3.13 and PyTorch CUDA 13.0)
Browser	Google Chrome 143+ recommended for best UI performance
Software	Microsoft Visual C++ Redistributable (for manual install)
Driver	Latest NVIDIA drivers

For GPU purchasing advice, check the GPU Buying Guide. If your hardware is modest, don't worry — install ComfyUI first, and if it runs too slowly, then upgrade.

Method 1: Official Portable Version (Recommended)

The official portable version is the quickest way to get ComfyUI running on Windows. It comes with a bundled Python 3.13 environment and PyTorch with CUDA 13.0 support.

Step 1: Download

Download the portable version that matches your GPU:

GPU Type	Download	Notes
NVIDIA (20 series+)	ComfyUI_windows_portable_nvidia.7z	Recommended for most users
NVIDIA (10 series & older)	ComfyUI_windows_portable_nvidia_cu126.7z	Uses PyTorch CUDA 12.6 + Python 3.12
AMD GPUs	ComfyUI_windows_portable_amd.7z	Experimental support (RDNA 3/3.5/4)
Intel GPUs	ComfyUI_windows_portable_intel.7z	Arc series with native torch.xpu

All downloads are available on the official releases page.

Step 2: Extract

Extract the downloaded .7z archive using 7-Zip (or Windows Explorer on recent Windows versions).

Trouble extracting? Right-click the file → Properties → Unblock, then try again.

After extraction, the folder structure should look like this:

ComfyUI_windows_portable
├── ComfyUI
│   ├── models
│   │   ├── checkpoints     ← Place checkpoint models here
│   │   ├── loras
│   │   ├── vae
│   │   ├── controlnet
│   │   └── ...
│   ├── custom_nodes
│   ├── user
│   └── ...
├── python_embeded          ← Bundled Python 3.13 environment
├── run_cpu.bat             ← For CPU or AMD GPUs
└── run_nvidia_gpu.bat      ← For NVIDIA GPUs

Step 3: Install a Model

ComfyUI needs at least one checkpoint model to generate images. Download a model like Anything V5 or choose from the Model Recommendations section, then place it in the models/checkpoints folder.

If you already have models from another UI (A1111 WebUI, etc.), see How to Share Models or edit extra_model_paths.yaml in the ComfyUI directory.

Step 4: Run ComfyUI

Double-click the appropriate batch file:

run_nvidia_gpu.bat — if you have an NVIDIA GPU
run_cpu.bat — if you have an AMD GPU or only a CPU

When the console shows:

Starting server
To see the GUI go to: http://127.0.0.1:8188

ComfyUI is running. Open your browser and navigate to http://127.0.0.1:8188.

Step 5: Generate Your First Image

Confirm that a checkpoint model is loaded in the Load Checkpoint node
Click Queue Prompt
Wait for the image to generate. If successful, congratulations — ComfyUI is installed and working!

If the page is blank or no default workflow loads, click Load Default from the menu in the lower-right corner.

If you see the page title but no GUI elements, your browser may not be compatible. Use the latest Chrome Browser.

Update your NVIDIA drivers if the portable version doesn't start.

Method 2: Manual Installation (Git + Venv)

For users who want full control, clone the repository and manage dependencies yourself.

Prerequisites

Git installed
Python 3.12 or 3.13 installed
Microsoft Visual C++ Redistributable installed

Steps

# 1. Clone the repository
git clone https://github.com/comfyanonymous/ComfyUI.git
cd ComfyUI

# 2. Create a virtual environment
python -m venv .venv

# 3. Activate the environment
.venv\Scripts\activate

# 4. Install PyTorch with CUDA support
pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu130

# 5. Install ComfyUI dependencies
pip install -r requirements.txt

# 6. Start ComfyUI
python main.py

The official repository has detailed manual installation instructions in the README.

Community Distribution: The Aaaki ComfyUI Launcher is a third-party launcher created by the community — not an official ComfyUI product. It's popular among Chinese-speaking users and supports Simplified Chinese, Traditional Chinese, English, and Japanese.

Other Resources

If you already use A1111 WebUI, see How to Share Models Between A1111 WebUI and ComfyUI
ComflowySpace — another third-party integration package