Installation options Choose the installation method that works best for you:
Desktop app
Windows portable
comfy-cli
Manual install
The easiest way to get started with ComfyUI. Download from the official website . Get the latest commits in a completely portable package. Download Download the latest Windows portable build Installation steps
Download and extract
Download the .7z file and extract it with 7-Zip or Windows Explorer (on recent Windows versions). If you have trouble extracting, right-click the file → Properties → Unblock
Add your models
Place your checkpoint files (.ckpt or .safetensors) in: ComfyUI\models\checkpoints
For larger models with multiple files, follow the model’s instructions to know which subfolder to use in ComfyUI\models\.
Run ComfyUI
Run the included executable to start ComfyUI. The portable build includes Python 3.13 and PyTorch CUDA 13.0. Update your NVIDIA drivers if it doesn’t start.
Alternative downloads Share models with other UIs Edit extra_model_paths.yaml.example in the ComfyUI directory, rename it to extra_model_paths.yaml, and configure your model paths. Install using the ComfyUI command-line interface. pip install comfy-cli comfy install
See the comfy-cli documentation for more details. Install from source for full control and customization. See the Manual installation section below. Manual installation Prerequisites Python version: Python 3.13 is recommended. Python 3.14 works but some custom nodes may have issues. If you have dependency issues on 3.13, try 3.12.
PyTorch version: PyTorch 2.4+ is supported. We recommend using the latest major version with the latest CUDA version (unless it’s less than 2 weeks old).
Installation steps
Clone the repository
git clone https://github.com/comfyanonymous/ComfyUI.git cd ComfyUI
Install PyTorch
Choose the installation command for your GPU: Stable version: pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu130
Nightly version (may have performance improvements): pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu130
If you get “Torch not compiled with CUDA enabled” error: pip uninstall torch # Then reinstall with the command above
Stable version (ROCm 7.1): pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm7.1
Nightly version (ROCm 7.2, may have performance improvements): pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/rocm7.2
RDNA 3 (RX 7000 series): pip install --pre torch torchvision torchaudio --index-url https://rocm.nightlies.amd.com/v2/gfx110X-all/
RDNA 3.5 (Strix Halo/Ryzen AI Max+ 365): pip install --pre torch torchvision torchaudio --index-url https://rocm.nightlies.amd.com/v2/gfx1151/
RDNA 4 (RX 9000 series): pip install --pre torch torchvision torchaudio --index-url https://rocm.nightlies.amd.com/v2/gfx120X-all/
Stable version: pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/xpu
Nightly version: pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/xpu
See the PyTorch XPU documentation for more information. Follow the Accelerated PyTorch training on Mac guide to install PyTorch nightly.
Install dependencies
pip install -r requirements.txt
Add your models
Place your models in the appropriate folders:
Checkpoints: models/checkpoints
VAE: models/vae
LoRAs: models/loras
Embeddings: models/embeddings
Running ComfyUI Basic usage ComfyUI will start and be accessible at http://127.0.0.1:8188 by default.
Command-line options Listen on specific address and port
CPU mode (slow)
Enable high-quality previews
Auto-launch browser
Enable ComfyUI-Manager
TLS/SSL support
python main.py --listen 0.0.0.0 --port 8188
AMD-specific options Unsupported cards
Performance tips
For AMD cards not officially supported by ROCm: RDNA2 or older (6700, 6600): HSA_OVERRIDE_GFX_VERSION = 10.3.0 python main.py
RDNA3 cards (7600): HSA_OVERRIDE_GFX_VERSION = 11.0.0 python main.py
Enable experimental memory-efficient attention: TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL = 1 python main.py --use-pytorch-cross-attention
Enable TunableOp (slow first run, faster subsequent runs): PYTORCH_TUNABLEOP_ENABLED = 1 python main.py
Installing ComfyUI-Manager ComfyUI-Manager allows you to easily install, update, and manage custom nodes.
Install manager dependencies
pip install -r manager_requirements.txt
Enable the manager
python main.py --enable-manager
Manager options Flag Description --enable-managerEnable ComfyUI-Manager --enable-manager-legacy-uiUse legacy manager UI (requires --enable-manager) --disable-manager-uiDisable manager UI but keep background features
High-quality previews By default, ComfyUI uses low-resolution previews. To enable high-quality TAESD previews:
Download TAESD models
Download the following files:
Place in VAE approx folder
Move the downloaded files to:
Launch with TAESD previews
python main.py --preview-method taesd
Next steps
Create your first workflow Now that ComfyUI is installed, learn how to create your first image generation workflow.