Skip to content

ComfyUI Install That Actually Works (No Crashes)

Most guides show how to install ComfyUI.

They don’t show how to install it so it doesn’t crash later.

This guide focuses on: - stability - predictable performance - avoiding common failures

0. Before You Start

Run system checks first:

df -h
nvidia-smi
free -h

You should have: - 50–100GB free disk - working GPU - 16GB+ RAM (recommended)

1. Install System Dependencies

sudo apt update
sudo apt install -y python3 python3-venv python3-pip git

2. Clone ComfyUI

Choose a stable location:

mkdir -p /opt/ai
cd /opt/ai

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

3. Create Virtual Environment

Never install globally.

python3 -m venv venv
source venv/bin/activate

4. Install Requirements

pip install --upgrade pip
pip install -r requirements.txt

5. Install PyTorch (GPU)

Check CUDA version:

nvidia-smi

Then install PyTorch.

Example (CUDA 12.x):

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

6. First Run

python main.py

Open in browser:

http://YOUR_SERVER_IP:8188

7. Create Models Directory

Do NOT mix models inside project.

Create separate storage:

mkdir -p /opt/models/{checkpoints,loras,vae,controlnet}

Instead of copying files, use symlinks:

ln -s /opt/models/checkpoints models/checkpoints
ln -s /opt/models/loras models/loras
ln -s /opt/models/vae models/vae

9. Run with GPU Optimization

Basic:

python main.py --listen 0.0.0.0

Low VRAM mode:

python main.py --lowvram

10. Common Failures (and Fixes)

Out of VRAM

Symptoms: - crash during generation - CUDA out of memory

Fix: 

python main.py --lowvram

Use smaller models.

Slow Performance

Check GPU:

nvidia-smi

If GPU usage is 0%: - wrong PyTorch build - CUDA mismatch

Module Errors

Fix:

pip install -r requirements.txt

Port Not Accessible

Run:

python main.py --listen 0.0.0.0

Check firewall.

11. Run in Background (Production)

Use screen:

screen -S comfyui
python main.py --listen 0.0.0.0

Detach:

Ctrl + A, then D

/opt/ai/ComfyUI
/opt/models
/opt/output

13. Why Most Installs Fail

Common mistakes:

  • installing without GPU support
  • running out of disk space
  • mixing models inside project
  • no virtual environment

14. Next Step

Now fix VRAM issues:

👉 VRAM Triage