---
title: "Installing ComfyUI on Linux"
date: 2025-01-15T15:30:00.000Z
author: Z.SHINCHVEN
tags: [ComfyUI, Linux, Stable Diffusion, AI, Installation Guide, NVIDIA, AMD, Intel Arc]
canonical: https://atlassc.net/2025/01/16/installing-comfyui-on-linux
---
## Introduction

ComfyUI is a highly flexible, node-based GUI and backend for Stable Diffusion. It's a favorite among power users for its performance and the ability to create complex generation workflows. Linux is often the best environment for running AI workloads due to its efficiency and developer-friendly tools.

This guide will walk you through installing ComfyUI on a Linux system, with specific instructions for **NVIDIA**, **AMD**, and **Intel Arc** GPUs.

## Prerequisites

Before starting, ensure your system is up to date and has the following installed:

*   **Git:** For cloning the repository.
*   **Python 3:** Version 3.12 or 3.13 is recommended. (3.14 works but is experimental).
*   **Virtual Environment (venv):** Highly recommended to keep dependencies isolated.

## Step 1: Clone the Repository

Open your terminal and clone the ComfyUI repository to your desired location:

```bash
git clone https://github.com/comfyanonymous/ComfyUI.git
cd ComfyUI
```

## Step 2: Set Up a Virtual Environment

Creating a virtual environment prevents conflicts with your system's Python packages.

```bash
# Create a virtual environment named 'venv'
python3 -m venv venv

# Activate the environment
source venv/bin/activate
```

*Note: You will need to run `source venv/bin/activate` every time you open a new terminal to run ComfyUI.*

## Step 3: Install PyTorch (GPU Specific)

This is the most critical step. You need to install the version of PyTorch that matches your hardware.

### Option A: NVIDIA GPUs (CUDA)

For NVIDIA users, CUDA is the standard.

*   **Stable Version (Recommended):**
    ```bash
    pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu130
    ```

*   **Nightly Version (For latest features):**
    ```bash
    pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu130
    ```

### Option B: AMD GPUs (ROCm)

AMD users should use ROCm. Support has improved significantly with recent updates.

*   **Stable Version (ROCm 6.4):**
    ```bash
    pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.4
    ```

*   **Nightly Version (ROCm 7.1+):**
    ```bash
    pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/rocm7.1
    ```

*   **For RDNA 3 / 4 (RX 7000/9000 Series) - Specific Nightlies:**
    If you have a modern card, use the specialized nightly builds for best performance:
    *   **RX 7000 Series:**
        ```bash
        pip install --pre torch torchvision torchaudio --index-url https://rocm.nightlies.amd.com/v2/gfx110X-dgpu/
        ```
    *   **RX 9000 Series:**
        ```bash
        pip install --pre torch torchvision torchaudio --index-url https://rocm.nightlies.amd.com/v2/gfx120X-all/
        ```

### Option C: Intel Arc GPUs (XPU)

Intel Arc users can use the native XPU support.

*   **Stable Version:**
    ```bash
    pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/xpu
    ```

## Step 4: Install Dependencies

Once PyTorch is installed, install the rest of the required Python packages:

```bash
pip install -r requirements.txt
```

## Step 5: Place Your Models

Before running the server, you need to add your Stable Diffusion models.

*   **Checkpoints:** Copy your `.ckpt` or `.safetensors` files to `ComfyUI/models/checkpoints/`.
*   **VAEs:** Copy VAE files to `ComfyUI/models/vae/`.

## Step 6: Run ComfyUI

You are now ready to launch ComfyUI!

```bash
python main.py
```

Access the interface by opening your browser to `http://127.0.0.1:8188`.

## Troubleshooting & Tips

*   **AMD RDNA 2 Issues:** If you have an RX 6000 series card (e.g., 6700 XT) and it crashes, try forcing the graphics version:
    ```bash
    HSA_OVERRIDE_GFX_VERSION=10.3.0 python main.py
    ```
*   **"Torch not compiled with CUDA enabled":** This means you installed the generic `torch` package instead of the CUDA version. Uninstall it (`pip uninstall torch`) and re-run the specific command from Step 3.
*   **Performance:**
    *   **AMD:** To enable experimental memory optimization: `TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1 python main.py --use-pytorch-cross-attention`
    *   **General:** If you have low VRAM, use `--lowvram` flag: `python main.py --lowvram`.

## Conclusion

Installing ComfyUI on Linux is a straightforward process that rewards you with a stable and high-performance environment for AI art generation. Whether you are Team Green, Red, or Blue, ComfyUI has you covered.

For the latest updates, keep an eye on the [official repository](https://github.com/comfyanonymous/ComfyUI). Happy prompting!