Sign In
homemodelsimagesvideospostsarticlesbountieschallengeseventsupdatesshop

ComfyUI for Linux & Windows

Updated: Apr 5, 2026

toollinuxwindowsportablevenvcomyfui

Download (27.87 KB)

Verified:

Other

Type

Other

Stats

24

0

Reviews

(8)

Published

Apr 3, 2026

Base Model

Other

Hash

AutoV2
670D30AB8F
Comfy Workflow Badge
Rvage's Avatar

Rvage

ComfyUI Installation Script for Linux & Windows

Automated installation scripts for ComfyUI on Linux and Windows systems that handle everything from Python environment setup to custom node installation. These scripts provide a streamlined, dependency-conflict-free installation process with full control over package versions and installation steps.

🌟 Features

  • Automated Environment Setup: Installs pyenv, Python, and creates an isolated virtual environment

  • Interactive Prompts: Collects ComfyUI version, frontend version, and alias name at runtime

  • Version Control: Pin specific versions of ComfyUI, PyTorch, NumPy, Transformers, and other critical packages

  • Frontend Pinning: Each alias/launcher auto-installs the correct frontend version on launch

  • GPU Acceleration: Supports CUDA 13.0, CUDA 12.8, CUDA 12.6, ROCm 7.1, and CPU-only installations

  • Performance Optimization: Includes optional Nunchaku, Flash Attention, and Sage Attention

  • Custom Node Collection: Automatically clones and configures 45+ popular custom nodes

  • Lowercase Cloning: Clones custom nodes with lowercase directory names to match ComfyUI-Manager convention

  • Selective Installation: Choose which components to install via interactive menu

  • Smart Dependency Management: Prevents version conflicts by enforcing package versions

  • Symlink Support: Centralize models and outputs across multiple installations

  • Multi-Install Support: Run multiple ComfyUI versions side-by-side with per-version aliases and launchers

  • Shell Integration: Adds user-chosen aliases (e.g., comfy2, comfy3) with non-destructive, additive handling

  • Multiple Shell Support: Auto-detects and configures bash, zsh, and fish shells

📋 Prerequisites

Linux

  • Operating System: Linux (Ubuntu, Debian, Fedora, Arch, openSUSE, etc.)

  • Permissions: Sudo access for installing system dependencies

  • Disk Space: ~10-20GB for full installation (varies with custom nodes)

  • GPU (optional): NVIDIA GPU with CUDA support for acceleration features

  • Git: For cloning repositories

Windows

  • Operating System: Windows 10 or Windows 11

  • PowerShell: Version 5.1+ (included with Windows 10/11) or PowerShell 7+

  • Disk Space: ~10-20GB for full installation

  • GPU (optional): NVIDIA GPU with CUDA support for acceleration features

  • Git: Git for Windows installed and in PATH (download)

  • Visual Studio Build Tools (optional): Required only if building packages from source (e.g., flash-attn)

🚀 Quick Start

Linux

# Download the script
git clone https://github.com/r-vage/ComfyUI-Installation-Script-for-Linux.git
cd ComfyUI-Installation-Script-for-Linux

# Make executable
chmod +x install_comfy_env.sh

# Run the script
./install_comfy_env.sh

Windows

# Download the script
git clone https://github.com/r-vage/ComfyUI-Installation-Script-for-Linux.git
cd ComfyUI-Installation-Script-for-Linux

# Allow script execution (if not already enabled)
Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser

# Run the script
.\install_comfy_env_win.ps1

Note: The Windows script uses directory junctions (similar to symlinks) which work without Administrator privileges. If you need true symlinks, run PowerShell as Administrator.

The script will display your configuration and prompt you to select which installation steps to run.

⚙️ Configuration

Before running the script, you can customize these variables at the top of install_comfy_env.sh (Linux) or install_comfy_env_win.ps1 (Windows):

Python Configuration

PYTHON_VERSION="3.12.10"          # Python version (3.10.x, 3.11.x, 3.12.x, 3.13.x, 3.14.x recommended)
VENV_PATH="/mnt/daten/AI/comfy_env"  # Virtual environment location
PYENV_ROOT="${PYENV_ROOT:-$HOME/.pyenv}"  # pyenv installation directory

Windows equivalent: $PYTHON_VERSION = "3.12.10" — venv path defaults to $BASE_PATH\comfy_env (e.g. D:\AI\comfy_env)

Recommended Python versions: 3.10.x, 3.11.x, 3.12.x, 3.13.x, 3.14.x have prebuilt wheels for PyTorch, Nunchaku, and Flash Attention. Other versions will require compilation from source.

PyTorch Configuration

PYTORCH_VERSION="2.9"             # PyTorch major.minor for wheel URLs
PYTORCH_FULL_VERSION="2.9.1+cu128" # Full version string
PYTORCH_INDEX_URL="https://download.pytorch.org/whl/cu128"  # Wheel repository

Available options:

  • CUDA 13.0: cu130 (latest)

  • CUDA 12.8: cu128 (default)

  • CUDA 12.6: cu126

  • ROCm 7.1: rocm7.1 (AMD GPUs)

  • CPU only: cpu (no GPU)

Example CUDA 13.0 configuration:

PYTORCH_FULL_VERSION="2.9.1+cu130"
PYTORCH_INDEX_URL="https://download.pytorch.org/whl/cu130"

Example ROCm 7.1 configuration (AMD GPUs):

PYTORCH_FULL_VERSION="2.9.1+rocm7.1"
PYTORCH_INDEX_URL="https://download.pytorch.org/whl/rocm7.1"

Example CPU configuration:

PYTORCH_FULL_VERSION="2.9.1+cpu"
PYTORCH_INDEX_URL="https://download.pytorch.org/whl/cpu"

Critical Package Versions

NUMPY_VERSION="2.2.6"             # NumPy version (2.2.x for PyTorch 2.9+)
TRANSFORMERS_VERSION="5.3.0"      # Transformers (5.x for Qwen3-VL/Mistral3)

These versions are enforced at the end of installation to override any conflicting dependencies from custom nodes.

ComfyUI Installation (Interactive)

When you run the script, it prompts for three values:

Enter values for this installation (press Enter for defaults):

  ComfyUI version [0.18.0]:     # → clones/checks out this tag (e.g., v0.18.2)
  Frontend version [1.41.21]:   # → pinned in aliases and launchers
  Launch alias [comfyui]:       # → shell alias name (e.g., comfy2, comfy3)

The folder name is derived automatically: version 0.18.2 → ComfyUI_0.18.2.

Static configuration (Python, PyTorch, NumPy, Transformers) stays at the top of the script and rarely changes.

Multiple ComfyUI Installations:
Simply run the script again with different values at the prompts:

# First run:   version 0.18.0, alias comfyui   → /mnt/data/AI/ComfyUI_0.18.0
# Second run:  version 0.18.2, alias comfy2     → /mnt/data/AI/ComfyUI_0.18.2
# Third run:   version 0.19.0, alias comfy3     → /mnt/data/AI/ComfyUI_0.19.0

Each installation gets its own alias and launcher script (start_comfy2.sh, start_comfy3.bat).
All installations share the same venv, models, output, and custom nodes via symlinks.
The envact alias is always shared (one venv for all installs).

CREATE_SYMLINKS=true                # Enable/disable symlink creation
USER_MODELS_PATH="/mnt/daten/AI/models"  # Centralized models directory
USER_OUTPUT_PATH="/mnt/daten/AI/output"  # Centralized output directory
USER_CUSTOM_NODES_PATH="/mnt/daten/AI/custom_nodes"  # Centralized custom_nodes directory

Windows equivalent: $CREATE_SYMLINKS = $true — paths default to D:\AI\models, D:\AI\output, D:\AI\custom_nodes (uses directory junctions instead of symlinks)

Symlinks allow you to:

  • Share models across multiple ComfyUI installations

  • Share custom nodes across multiple ComfyUI installations (useful when testing different ComfyUI versions)

  • Store models on a different drive/partition

  • Keep outputs in a centralized location

Set CREATE_SYMLINKS=false to use default ComfyUI/models, ComfyUI/output, and ComfyUI/custom_nodes directories.

Tip — directories outside BASE_PATH:
By default, USER_MODELS_PATH, USER_OUTPUT_PATH, and USER_CUSTOM_NODES_PATH are derived from BASE_PATH. You can point any of them to a completely different location — for example a larger disk — by editing the "Derived Paths" section in the script:

Linux:

USER_MODELS_PATH="/mnt/ssd2/models"           # Models on a fast SSD
USER_OUTPUT_PATH="/home/$USER/comfyui_output"  # Output in home directory

Windows:

$USER_MODELS_PATH = "E:\models"                         # Models on drive E:
$USER_OUTPUT_PATH = "C:\Users\$env:USERNAME\comfyui_out" # Output in user profile

The script creates the target directories automatically and symlinks/junctions them into the ComfyUI tree.

Optional Features

INSTALL_NUNCHAKU=true               # Nunchaku acceleration (requires NVIDIA GPU)

Set to false to skip Nunchaku installation if you don't have an NVIDIA GPU or don't need this optimization.

📦 What Gets Installed

Installation Steps

The script is divided into 11 steps that you can run selectively:

  1. Python Environment - pyenv, Python version, virtual environment

  2. PyTorch - PyTorch, torchvision, torchaudio with CUDA/CPU support

  3. Nunchaku - Acceleration library (optional, NVIDIA GPU only)

  4. Face Recognition - facexlib, insightface, onnxruntime-gpu, facenet_pytorch

  5. ComfyUI Core - ComfyUI base installation and requirements

  6. Custom Nodes - 45+ popular custom nodes (see list below)

  7. Custom Node Dependencies - Install requirements for all custom nodes

  8. Performance Libraries - llama-cpp-python, flash-attn, sageattention

  9. Upgrade/Pin Packages - Upgrade specific packages to latest compatible versions

  10. Enforce Versions - Force exact versions of PyTorch, NumPy, Transformers, ComfyUI Frontend

  11. Shell Aliases - Add user-chosen launch alias and envact alias to shell config

Custom Nodes Included (45 nodes)

Core Extensions

UI & Workflow Tools

Model Support & Optimization

Sampling & Scheduling

ControlNet & Advanced Control

Image Processing & Effects

Specialized Models

Video Processing

🎯 Usage

Interactive Installation

Run the script and select steps interactively:

./install_comfy_env.sh

You'll see a menu:

Select installation steps (numbers, ranges, or 'a' for all — e.g., 6-10 or 1 5 6-8):
  1) Python environment (pyenv + venv)
  2) PyTorch and base dependencies
  ...
  a) All steps (default)

Your selection [a]:

Examples:

  • a - Install everything (default)

  • 1 2 5 - Only setup Python, PyTorch, and ComfyUI core

  • 6-7 - Only clone and setup custom nodes (if you already have ComfyUI)

  • 5 9-11 - ComfyUI core + upgrade/pin/aliases

  • 10 - Only enforce package versions (useful after manual package changes)

Running ComfyUI

After installation, you have several options (examples assume alias comfy2 and version 0.18.2):

Linux

Option 1: Launcher script (recommended)

/mnt/data/AI/start_comfy2.sh

Option 2: Shell alias (if Step 11 was run)

comfy2    # Auto-installs correct frontend, activates env, launches ComfyUI

Option 3: Manual

source /mnt/data/AI/comfy_env/bin/activate
cd /mnt/data/AI/ComfyUI_0.18.2
python main.py

Windows

Option 1: Double-click launcher

D:\AI\start_comfy2.bat

Option 2: PowerShell launcher

D:\AI\start_comfy2.ps1

Option 3: PowerShell alias (if Step 11 was run, after reloading profile)

comfy2

Option 4: Manual

& "D:\AI\comfy_env\Scripts\Activate.ps1"
cd "D:\AI\ComfyUI_0.18.2"
python main.py

Each alias and launcher automatically pins the correct frontend version via uv pip install -q comfyui-frontend-package==VERSION (instant no-op when already installed).

Activating Virtual Environment Only

# Shared across all installations (Step 11)
envact

# Or manually
source /mnt/data/AI/comfy_env/bin/activate

🔧 Customizing Custom Nodes

To add/remove custom nodes, edit the [6/10] Clone Custom Nodes section around line 550:

# Add your custom node:
clone_if_missing "https://github.com/yourusername/your-custom-node.git"

# Remove a node by commenting it out or deleting the line:
# clone_if_missing "https://github.com/some/node-you-dont-want.git"

Then run:

./install_comfy_env.sh

And select steps 6-7 to update custom nodes and their dependencies.

🐛 Troubleshooting

pyenv Installation Fails

If pyenv dependencies fail to install, you may need to manually install build dependencies for your distribution:

Ubuntu/Debian:

sudo apt-get update
sudo apt-get install build-essential libssl-dev zlib1g-dev \
libbz2-dev libreadline-dev libsqlite3-dev curl git \
libncursesw5-dev xz-utils tk-dev libxml2-dev libxmlsec1-dev libffi-dev liblzma-dev

Fedora:

sudo dnf install make gcc patch zlib-devel bzip2 bzip2-devel readline-devel sqlite sqlite-devel openssl-devel tk-devel libffi-devel xz-devel libuuid-devel gdbm-libs libnsl2

Nunchaku or Flash Attention Installation Fails

These require:

  • NVIDIA GPU

  • CUDA 12.6+ or compatible version

  • Prebuilt wheels for your Python version

If you don't have an NVIDIA GPU, set INSTALL_NUNCHAKU=false and skip step 3.

Custom Node Dependencies Conflict

If a custom node installs incompatible package versions, run step 10 to enforce configured versions:

./install_comfy_env.sh  # Select step 10 only

Shell Aliases Not Working

Linux

After step 11, you need to reload your shell configuration:

# For bash
source ~/.bashrc

# For zsh
source ~/.zshrc

# For fish
source ~/.config/fish/config.fish

# Or just restart your terminal

Windows

After step 11, reload your PowerShell profile:

. $PROFILE

# Or just restart PowerShell

Windows: Execution Policy Error

If you get a script execution error:

Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser

Windows: pyenv-win Not Found After Installation

Close and reopen PowerShell. If still not found, ensure these are in your PATH:

  • %USERPROFILE%\.pyenv\pyenv-win\bin

  • %USERPROFILE%\.pyenv\pyenv-win\shims

Windows: Flash Attention Not Available

Flash Attention rarely has prebuilt Windows wheels. You can either:

  • Skip it (it's optional — ComfyUI works fine without it)

  • Install Visual Studio Build Tools and compile from source

  • Use Sage Attention as an alternative

Out of Disk Space

The full installation requires 10-20GB. To reduce space:

  • Skip custom nodes (don't run step 6-7)

  • Remove unused custom nodes manually from custom_nodes/ directory

  • Use symlinks to store models on a different drive

Happy ComfyUI-ing! 🎨

Version 0.8.2

  • since im not using windows anymore there will be no further updates for the portable versions 🤷🏻.

  • pyTorch 2.9+cu128

  • python: 3.12.10

  • Frontend: 1.36.13

  • removed crystools -> replaced with crystools-monitor only

  • Even though some API imports have been fixed, the deprecated messages are still displayed.

  • the reason for adding the forked versions instead of the originals is that some of the nodes are currently not working with the latest comfyui version.

Additional Installed Libs

  • triton 3.51.1

  • llama_cpp_python 0.3.16

  • flash attention 2 (v2.8.3+torch2.9+cu128-cp312)

  • nunchaku-1.0.2+torch2.9-cp312-cp312

  • sage attention 2.2.0

  • transformers==4.56.2

  • insightface, facexlib, facenet_pytorch, comfy-cli, bitsandbytes, tensorflow, hf_xet, requests, pilgram, tf-keras, setuptools_scm, qwen-vl-utils, huggingface-hub, accelerate, Pillow, opencv-python, safetensors, psutil, PyYAML, aiohttp, av, gguf, ultralytics, pytorch-extension

  • python headers for torch compile/triton

Installed Custom Nodes

  • comfy-mtb

  • comfyui_birefnet_ll

  • comfyui_controlnet_aux

  • ComfyUI_Eclipse 1.2.1

  • comfyui_essentials

  • comfyui_essentials_mb

  • comfyui_layerstyle

  • ComfyUI_LayerStyle_Advance

  • ComfyUI_Patches_ll (forked and fixed Imports)

  • comfyui_pulid_flux_ll

  • comfyui_ttp_toolset

  • comfyui-advanced-controlnet

  • ComfyUI-Crystools-MonitorOnly (forked and fixed small error in browser log)

  • ComfyUI-Custom-Scripts (forked and fixed API Imports)

  • comfyui-detail-daemon

  • comfyui-easy-use

  • comfyui-florence2

  • comfyui-frame-interpolation

  • ComfyUI-GGUF

  • ComfyUI-GIMM-VFI

  • comfyui-impact-pack

  • comfyui-kjnodes

  • ComfyUI-Manager (removed the anoying fetch msg output)

  • ComfyUI-nunchaku (forked and fixed Imports)

  • ComfyUI-ReActor

  • comfyui-supir

  • ComfyUI-TeaCache (forked and fixed Imports)

  • comfyui-videohelpersuite

  • ComfyUI-WanVideoWrapper

  • raffle

  • RES4LYF (forked and partially fixed (community pull requests))

  • rgthree-comfy (forked and fixed API Imports)

  • sd-dynamic-thresholding

  • was-ns

Version 0.3.75

(You will see some warnings from various nodes because the authors of comfyui are changing the API. -> [DEPRECATION WARNING])

Additional Installed Libs

  • triton 3.51.1

  • llama_cpp_python 0.3.16

  • flash attention 2 (v2.8.3+torch2.9+cu128-cp312)

  • nunchaku-1.0.2+torch2.9-cp312-cp312

  • sage attention 2.2.0

  • transformers==4.56.2

  • facenet_pytorch, pytorch-extension, facexlib, insightface, comfy-cli, bitsandbytes, tensorflow, hf_xet, requests, pilgram, tf-keras, nvidia-ml-py

  • python headers for torch compile/triton

Installed Custom Nodes

  • comfy-mtb

  • comfyui_birefnet_ll

  • comfyui_controlnet_aux

  • ComfyUI_Eclipse (1.0.60)

  • comfyui_essentials

  • ComfyUI_essentials_mb

  • comfyui_layerstyle

  • ComfyUI_LayerStyle_Advance

  • comfyui_patches_ll

  • comfyui_pulid_flux_ll

  • comfyui_ttp_toolset

  • comfyui-advanced-controlnet

  • ComfyUI-Crystools

  • comfyui-custom-scripts

  • comfyui-depthanythingv2

  • comfyui-detail-daemon

  • comfyui-easy-use

  • comfyui-florence2

  • comfyui-frame-interpolation

  • ComfyUI-GGUF

  • ComfyUI-GIMM-VFI

  • comfyui-impact-pack

  • comfyui-impact-subpack

  • comfyui-inspire-pack

  • comfyui-jdcn

  • comfyui-kjnodes

  • ComfyUI-Manager

  • comfyui-multigpu

  • ComfyUI-nunchaku

  • comfyui-reactor

  • comfyui-supir

  • comfyui-videohelpersuite

  • ComfyUI-WanMoeKSampler

  • ComfyUI-WanVideoWrapper

  • facerestore_cf

  • raffle

  • RES4LYF

  • rgthree-comfy

  • sd-dynamic-thresholding

  • sdxl_prompt_styler

  • teacache

  • was-ns

  • whiterabbit