Install ComfyUI and CatVTON on WSL (Windows Guide)

Install ComfyUI and CatVTON on WSL: Step-by-Step Guide

Some ComfyUI nodes, like CatVTON with Detectron2 and DensePose, are much easier to install on Linux than on Windows. WSL 2 gives you a real Ubuntu inside Windows with access to your NVIDIA GPU. This guide installs WSL, CUDA, Miniconda, ComfyUI and CatVTON step by step.

First published February 2025 on the old PixelAI Labs blog. Refreshed in October 2026: the steps below are checked, the links point to the current project pages, and newer options are mentioned where they help.

Requirements

Windows 10 (version 1903 or later) or Windows 11, 64-bit
Virtualization turned on in the BIOS/UEFI
An NVIDIA GPU with a current Windows driver

1. Install WSL 2 and Ubuntu

Turn on Windows Subsystem for Linux and Virtual Machine Platform in Windows Features, restart, then in PowerShell:

bash

wsl --install
wsl --set-default-version 2
wsl -l -v

If Ubuntu shows version 1, upgrade it:

bash

wsl --set-version ubuntu 2

2. Set up Ubuntu

bash

sudo apt update
sudo apt upgrade
sudo apt install -y build-essential

Install CUDA 12.1 with GCC 11:

bash

wget https://developer.download.nvidia.com/compute/cuda/12.1.0/local_installers/cuda_12.1.0_530.30.02_linux.run
sudo apt install gcc-11 g++-11
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-11 100
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-11 100
sudo sh cuda_12.1.0_530.30.02_linux.run

Add CUDA to your environment (nano ~/.bashrc), then reload it and check:

bash

export PATH=/usr/local/cuda-12.1/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-12.1/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
source ~/.bashrc
nvcc --version

Install cuDNN for CUDA 12 (Ubuntu 24.04):

bash

wget https://developer.download.nvidia.com/compute/cudnn/9.7.1/local_installers/cudnn-local-repo-ubuntu2404-9.7.1_1.0-1_amd64.deb
sudo dpkg -i cudnn-local-repo-ubuntu2404-9.7.1_1.0-1_amd64.deb
sudo cp /var/cudnn-local-repo-ubuntu2404-9.7.1/cudnn-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cudnn-cuda-12

3. Install Miniconda

bash

wget https://repo.anaconda.com/miniconda/Miniconda3-py311_24.4.0-0-Linux-x86_64.sh
bash Miniconda3-py311_24.4.0-0-Linux-x86_64.sh
source ~/.bashrc

Building an AI influencer? The UltraRealistic AI Influencer Pro course shows the complete system with today's models: design one character, train her own LoRA, and make KREA 2 photos and MiniMax H3 talking videos, on the cloud or your own GPU.

4. Install ComfyUI and CatVTON

bash

mkdir comfy && cd comfy
git clone https://github.com/Comfy-Org/ComfyUI.git
cd ComfyUI/custom_nodes
git clone https://github.com/Comfy-Org/ComfyUI-Manager.git comfyui-manager
git clone https://github.com/pzc163/Comfyui-CatVTON.git
cd ..

Create the environment from the main ComfyUI folder:

bash

conda create --name comfyui python=3.11
conda activate comfyui
conda install pytorch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 pytorch-cuda=12.1 -c pytorch -c nvidia
pip install -r requirements.txt

5. CatVTON dependencies

bash

pip install git+https://github.com/facebookresearch/detectron2.git@v0.6
pip install git+https://github.com/facebookresearch/detectron2.git@v0.6#subdirectory=projects/DensePose
pip install accelerate diffusers fvcore huggingface_hub hydra-core omegaconf pycocotools scikit-image yacs
pip install ninja==1.11.1.1 iopath==0.1.9 setuptools==58.2.0 Pillow==9.5.0

If ComfyUI later complains about libstdc++, link the system library into your conda environment (adjust the path if your environment isn't called comfyui):

bash

sudo apt install --only-upgrade libstdc++6
rm -rf ~/.cache/torch_extensions/
cd ~/miniconda3/envs/comfyui/lib
ln -sf /usr/lib/x86_64-linux-gnu/libstdc++.so.6 libstdc++.so.6

6. Run ComfyUI

bash

conda activate comfyui
cd ~/comfy/ComfyUI
python main.py

Open the address it prints in your Windows browser. The first run downloads CatVTON's models; the basic workflow is in custom_nodes/Comfyui-CatVTON/workflow.

FAQ

Why use WSL instead of Windows?

Nodes with Linux-first dependencies such as Detectron2 install cleanly on Ubuntu, and WSL 2 still uses your NVIDIA GPU.

Does WSL slow down generation?

GPU speed is about the same. Reading models from Windows drives (/mnt/c) is slower, so keep models inside the Linux file system.

CatVTON broke after I installed other nodes. What now?

Activate the conda environment again and reinstall CatVTON's dependencies from step 5.

Take it further

Once WSL is set up, most Linux-only nodes become easy to install. If you want a ready-made, tested system instead of assembling one yourself, the UltraRealistic AI Influencer Pro course walks through character design, LoRA training, KREA 2 photos and MiniMax H3 videos step by step, and the InstantClone AI Influencer Suite creates consistent images and motion videos from one character image. More free guides are on the PixelAI Labs blog.