How to Install ComfyUI on Windows, macOS, and Linux

Release check — 24 October 2026: the official GitHub latest release points to ComfyUI 0.42.1, released on 19 October. This patch fixes crashes some users encountered with the MOSS-Image template. The preceding 0.42.0 release added MOSS-Image support and the Krea 3 Image partner node. Follow our MOSS-Image setup and troubleshooting guide for that workflow. GitHub master commits and frontend/Desktop version numbers are separate from the core release tag; check your installed channel before updating.

ComfyUI installation can be simple or highly controlled depending on how you plan to use it. The official project now offers Comfy Desktop for supported operating systems, a Windows portable package, manual installation and command-line tooling.

This guide focuses on a clean setup that is easy to update and troubleshoot later.

ComfyUI installation: choose your operating system

Official ComfyUI downloads

Choose the official Desktop download for Windows, macOS or Linux, the Windows portable package, or the manual installation guide. These upstream pages provide current packages and hardware requirements; avoid downloading executables from unofficial mirrors.

How to install ComfyUI on Linux

For a Linux desktop, choose the official .deb or AppImage route above. The following commands cover the alternative manual installation path.

Linux users: install Git and a supported Python version, create and activate a virtual environment, clone the official repository, then install the PyTorch build appropriate for your CPU/GPU before the remaining requirements. Check the current official manual guide for backend-specific commands; CUDA, ROCm and CPU-only installations are not interchangeable.

Linux manual install: Ubuntu/Debian example

For a clean Linux installation, use a dedicated Python virtual environment. Install Git, Python and the virtual-environment package from your distribution first. Replace python3 with a supported Python interpreter if your distribution’s default is incompatible with the current ComfyUI release.

sudo apt update
sudo apt install -y git python3 python3-venv python3-pip
git clone https://github.com/Comfy-Org/ComfyUI.git
cd ComfyUI
python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip

Next, install PyTorch for your actual hardware using the official PyTorch selector (NVIDIA CUDA, supported AMD ROCm, or CPU). The same installation command does not work across all GPUs. Once PyTorch is installed, finish with:

python -m pip install -r requirements.txt
python -c "import torch; print('PyTorch:', torch.__version__); print('CUDA available:', torch.cuda.is_available())"
python main.py

Open http://127.0.0.1:8188 locally. A False CUDA result is expected on CPU-only and non-CUDA backends; it does not by itself prove a broken install. For ROCm or remote-server access, follow the official manual installation guide. Do not expose ComfyUI directly to the public internet without access controls.

For command-line installation and dependency management, the official Comfy CLI guide provides an alternative to the methods above.

ComfyUI system requirements (Desktop: Windows, macOS, Linux)

The official Comfy Desktop documentation currently lists support for Windows 10+, macOS 13+ on Apple Silicon, and Debian-based Linux. Each standalone installation requires several gigabytes of disk space, and the docs recommend more RAM than the bare minimum for comfortable use.

Remember that ComfyUI itself is only part of the storage requirement. Model checkpoints can add many more gigabytes.

Install with Comfy Desktop

For most beginners, this is the cleanest route:

1. download Comfy Desktop from the official ComfyUI site;
2. install the launcher for your operating system;
3. create a ComfyUI instance;
4. let the app manage the Python environment and dependencies;
5. launch the instance and run a known-good template before installing custom nodes.

The advantage is isolation: each instance can keep its own ComfyUI version, settings and custom nodes.

Install ComfyUI on Windows: portable package

The official Windows portable package is a self-contained .7z archive that you extract with 7-Zip. It ships with its own embedded Python environment (python_embeded), so it does not touch your system Python. Pick the package that matches your GPU:

• NVIDIA (RTX series): ComfyUI_windows_portable_nvidia.7z, built on CUDA 13.0 and Python 3.13.
• Older NVIDIA (10 series and earlier): ComfyUI_windows_portable_nvidia_cu126.7z, built on CUDA 12.6 and Python 3.12.
• AMD: ComfyUI_windows_portable_amd.7z.
• Intel: ComfyUI_windows_portable_intel.7z.

After extracting, double-click run_nvidia_gpu.bat to start ComfyUI (or run_cpu.bat for CPU only; the AMD and Intel packages include their own launchers). When the console prints the local address, http://127.0.0.1:8188, open it in your browser and keep the console window open, because closing it stops ComfyUI.

To update, run update_comfyui_stable.bat from the update folder for the latest stable version, or update_comfyui.bat for the latest commit. File names and CUDA versions change between releases, so check the official Windows portable guide for the current packages.

Manual installation: choose the right GPU backend

Use the virtual-environment setup and repository commands in the Linux/manual walkthrough above. The same principles apply on other supported systems: activate the correct environment, install the hardware-specific PyTorch build, then install ComfyUI requirements. Follow the official manual guide for operating-system differences.

NVIDIA

NVIDIA is the most broadly supported path for many ComfyUI workflows. Install a PyTorch build compatible with your current driver/CUDA environment, then install the ComfyUI requirements.

AMD

AMD support uses ROCm on supported systems. Hardware and operating-system support changes over time, so verify the current ROCm matrix before committing to a specific build.

How to install ComfyUI on Mac (macOS, Apple Silicon)

The official Comfy Desktop app for macOS requires macOS 13 (Ventura) or later on Apple Silicon (M1 or later), with roughly 4.85 GB of disk space per installation. Intel Macs are not listed in the official requirements.

1. Open the downloaded .dmg file.
2. Drag Comfy Desktop into the Applications folder.
3. Launch it from Launchpad or Spotlight.
4. If macOS shows a security warning on first launch, go to System Settings → Privacy & Security and click Open Anyway.

Prefer manual control? The official manual guide uses a Miniconda environment, clones the repository, installs PyTorch from the nightly channel with conda install pytorch-nightly::pytorch torchvision torchaudio -c pytorch-nightly, then runs pip install -r requirements.txt and python main.py. Check the current manual guide before copying commands, because nightly builds change.

Where do models go?

ComfyUI uses dedicated model directories such as:

• models/checkpoints
• models/loras
• models/vae
• models/controlnet
• models/diffusion_models

Different model families may require additional folders for text encoders, CLIP Vision or other components.

Share models between multiple interfaces

If you already have a large model library, you do not need to duplicate every file. ComfyUI supports extra model paths through extra_model_paths.yaml. This can point multiple ComfyUI instances, or ComfyUI plus another interface, at one central model folder.

Install custom nodes carefully

Do not install dozens of custom-node packs before confirming the base installation works. First run a standard workflow. Then add one extension at a time, preferably through a known source or ComfyUI-Manager.

Custom nodes execute code inside your environment, so treat them as software dependencies.

Optional Linux hardware optimization: owners of supported RTX 30 cards can read our gpu-unlock compatibility and benchmark analysis. It is a separate kernel-level project, not a requirement for installing ComfyUI, and reported compute gains are larger than the measured end-to-end gains in some image workflows.

How to update ComfyUI safely

Back up workflow JSON files, custom nodes and configuration before upgrading. Check the official release notes and test a simple workflow before relying on third-party extensions.

• Comfy Desktop: use the app’s built-in update controls rather than modifying its managed Python environment manually.
• Windows portable: use the update scripts shipped with your package, described in the portable section above.
• Manual Git installation: activate the existing environment, inspect local changes, update the checkout and reinstall its requirements.

cd ComfyUI
git status
git pull --ff-only
python -m pip install -r requirements.txt
python main.py

If local changes prevent a fast-forward update, review and back them up before continuing. Do not use these Git commands inside a managed Desktop installation. Record the working ComfyUI and custom-node versions so you can restore them if an upgrade causes a regression.

First-run checklist

1. launch ComfyUI without errors;
2. open the browser interface;
3. load a simple official/template workflow;
4. confirm the required model files are detected;
5. generate one output successfully;
6. save the workflow;
7. only then add custom nodes or complex video pipelines.

Common problems

Fix: [ComfyUI-Manager] PyTorch is not installed

This error often means ComfyUI-Manager is running with a different Python interpreter from the one where PyTorch was installed. Check the interpreter used to launch ComfyUI before reinstalling dependencies.

Manual Linux/macOS setup: activate the same virtual environment as ComfyUI:

cd ComfyUI
source .venv/bin/activate
python -c "import sys; print(sys.executable)"
python -m pip show torch
python -c "import torch; print(torch.__version__)"

If pip show torch returns nothing or importing torch fails, install the correct PyTorch build in that environment using the official selector, then restart ComfyUI. For Windows portable, open Command Prompt in the portable package folder and use the bundled interpreter (adjust the folder name if your package differs):

.\python_embeded\python.exe -m pip show torch
.\python_embeded\python.exe -c "import torch; print(torch.__version__)"

Do not run a system-wide pip install torch to repair a portable installation. If torch is missing, choose the hardware-compatible build from the PyTorch installation selector and install it with the same bundled interpreter. For Comfy Desktop, use the app-managed instance environment rather than assuming the portable Python path applies. If the error persists, inspect the complete startup log and any recently installed custom nodes.

Model not found

Check the folder, file type and extra-model-path configuration, then restart ComfyUI after changing paths.

CUDA or ROCm errors

These usually point to a mismatch between drivers, PyTorch build and hardware support. Verify the current backend installation rather than repeatedly reinstalling ComfyUI itself.

Out of memory

Lower resolution, batch size or frame count, use an appropriate precision/quantized model, or use supported offloading options. Video models are particularly memory intensive.

ComfyUI-Manager missing after an update?

On current manual installations, the official repository documents installing manager_requirements.txt and launching with python main.py --enable-manager. For Windows Portable, use its bundled Python interpreter and add --enable-manager to the appropriate launcher; do not install manager dependencies into the system Python. Desktop users should follow the application’s integrated manager/update controls.

Missing custom nodes

A downloaded workflow may depend on third-party nodes. Identify the missing nodes, install only those you actually need and restart the application.

Custom-node UI freezes: reported version-specific issues

Known user report, not a confirmed universal regression. A GitHub issue opened October 16, 2026 describes the ComfyUI canvas becoming non-interactive when loading a MiniMax H3 workflow using ComfyUI-DaSiWa-Nodes on ComfyUI 0.40.0 and 0.41.0. The reporter also saw warnings about deprecated frontend APIs (/scripts/ui.js and older extension components). This does not establish that every workflow or every custom node is affected.

What to do if a workflow freezes the canvas:

1. Back up the workflow JSON and record the ComfyUI version, frontend version and installed custom-node versions.
2. Restart ComfyUI and try loading a simple built-in workflow. If that works, isolate the custom extension or imported workflow rather than reinstalling everything.
3. Check the browser developer console and ComfyUI startup log for deprecated frontend API warnings or JavaScript errors. Update the affected extension only after checking its release notes and compatibility.
4. If the problem began immediately after an update, restore your previously working ComfyUI/custom-node combination from a backup while following the linked issue for a verified fix. Do not assume downgrading ComfyUI alone resolves a third-party extension conflict.

Different issue: MiniMax H3 GPU memory errors. A separate October 15 report describes an AMD Strix Halo/ROCm Dynamic VRAM out-of-memory failure. The reporter found --disable-dynamic-vram worked on that system. This is hardware- and workflow-specific and should not be presented as a universal fix for UI freezes.

Next steps

Once the installation is stable, continue with the main ComfyUI guide and ComfyUI Video Workflows.

No GPU? Run ComfyUI in the Cloud Instead

If your computer does not have a suitable NVIDIA GPU, or you want to run larger image and video workflows without managing CUDA, Python and local model storage, the official hosted option is Comfy Cloud. It runs ComfyUI workflows on remote GPUs while you work from the browser.

Read our ComfyUI Cloud guide for setup, pricing and the current free trial, or see Comfy Cloud vs Local GPU to compare cost, VRAM, custom nodes, privacy and workflow control before choosing.

ComfyUI install FAQ

What is the easiest way to set up ComfyUI?
For most supported desktop users, Comfy Desktop is the simplest starting point because it manages the application environment and dependencies. Windows users can also choose the portable package, while advanced users and servers may prefer a manual installation.

Where should I download ComfyUI?
Use the official ComfyUI documentation or the official ComfyUI GitHub repository linked in the Sources section below. Avoid unofficial installers when you cannot verify what they contain.

How do I install ComfyUI on Windows?
Windows users can choose Comfy Desktop, the official portable package, or a manual Python installation. The portable option is useful when you want the installation contained in its own folder.

Can I run ComfyUI on a Mac?
Comfy Desktop supports current Apple Silicon Macs on supported macOS versions. If a local workflow is too demanding for your hardware, a hosted GPU option can run the workflow remotely.

What are the minimum requirements for ComfyUI?
Requirements depend heavily on the model and workflow. The application itself can run on relatively modest systems, but modern image and especially video models can require substantially more RAM, VRAM and storage. Check the current official requirements and the model requirements before choosing hardware.

Sources

• Comfy Desktop official documentation
• Comfy Desktop for macOS
• ComfyUI manual installation documentation
• ComfyUI Windows portable documentation
• ComfyUI official repository