ComfyUI

ComfyUI is a node-based user interface for AI image generation. Instead of a traditional form-based layout, ComfyUI uses a visual graph editor where you connect processing nodes to build custom generation workflows. This approach offers significantly more flexibility than AUTOMATIC1111's WebUI, is generally faster due to built-in workflow optimizations, and supports advanced pipelines like multi-model generation, cascaded refinement, and custom node logic.

Prerequisites

Before installing ComfyUI, make sure you have the following:

Debian 12 (Bookworm) or later -- 64-bit x86_64
Python 3.10 or later -- Required runtime
NVIDIA GPU with 4 GB+ VRAM -- 8 GB+ recommended for SDXL models
15 GB+ free disk space -- For the application and model files
Git -- For cloning the repository

TIP

ComfyUI uses less VRAM than AUTOMATIC1111 by default, making it a better choice for users with limited GPU memory.

Installation

Method 1: comfy-cli (Recommended)

The comfy-cli tool automates the entire installation process.

bash

# Install comfy-cli
pip install comfy-cli

# Install ComfyUI (interactive setup)
comfy install

# Launch ComfyUI
comfy launch

During installation, comfy-cli will ask you to choose:

Installation directory
PyTorch version (select CUDA for NVIDIA GPUs)
Whether to download a default model

Method 2: Manual Installation

bash

# Install system dependencies
sudo apt update
sudo apt install -y python3 python3-pip python3-venv git wget

# Clone the ComfyUI repository
git clone https://github.com/comfyanonymous/ComfyUI.git
cd ComfyUI

# Create and activate a Python virtual environment
python3 -m venv venv
source venv/bin/activate

# Install PyTorch with CUDA support (for NVIDIA GPUs)
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124

# Install ComfyUI dependencies
pip install -r requirements.txt

Verify PyTorch CUDA Setup

bash

# Activate the virtual environment
cd ComfyUI
source venv/bin/activate

# Verify PyTorch can see the GPU
python3 -c "import torch; print('CUDA available:', torch.cuda.is_available()); print('GPU:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"

Download Models

You need at least one Stable Diffusion checkpoint to generate images.

bash

# Navigate to the checkpoints directory
cd ComfyUI/models/checkpoints

# Download Stable Diffusion XL base model
wget "https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors"

# Download a VAE for SDXL (optional but recommended)
cd ../vae
wget "https://huggingface.co/stabilityai/sdxl-vae/resolve/main/sdxl_vae.safetensors"

Configuration

Launch Options

bash

# Start ComfyUI (default: http://127.0.0.1:8188)
cd ComfyUI
source venv/bin/activate
python3 main.py

# Listen on all interfaces (accessible from other devices)
python3 main.py --listen 0.0.0.0

# Use a custom port
python3 main.py --port 8188

# Low VRAM mode (for GPUs with 4 GB or less)
python3 main.py --lowvram

# Use CPU only (very slow, for testing only)
python3 main.py --cpu

Argument	Description
`--listen`	Bind address (default: `127.0.0.1`)
`--port`	Port number (default: `8188`)
`--lowvram`	Reduce VRAM usage for limited GPUs
`--highvram`	Keep models in VRAM for faster generation
`--cpu`	Run on CPU only
`--preview-method auto`	Enable live preview during generation
`--use-pytorch-cross-attention`	Use PyTorch's built-in cross-attention

Model Directory Structure

ComfyUI/models/
  checkpoints/      # Main Stable Diffusion models (.safetensors)
  loras/            # LoRA models
  vae/              # VAE models
  controlnet/       # ControlNet models
  upscale_models/   # Upscaler models (ESRGAN, etc.)
  clip/             # CLIP text encoder models
  embeddings/       # Textual inversion embeddings

If you already have models from AUTOMATIC1111, you can share them instead of downloading duplicates.

bash

# Edit the extra model paths configuration
cp ComfyUI/extra_model_paths.yaml.example ComfyUI/extra_model_paths.yaml

Edit extra_model_paths.yaml:

yaml

# extra_model_paths.yaml
a111:
  base_path: /home/user/stable-diffusion-webui

  checkpoints: models/Stable-diffusion
  configs: models/Stable-diffusion
  vae: models/VAE
  loras: |
    models/Lora
    models/LyCORIS
  upscale_models: |
    models/ESRGAN
    models/RealESRGAN
  embeddings: embeddings
  controlnet: models/ControlNet

Usage

Basic Workflow

Open http://localhost:8188 in your browser
The default workflow appears with basic txt2img nodes
Click on the KSampler node to set sampling parameters
Click on CLIP Text Encode nodes to enter your positive and negative prompts
Click Queue Prompt to generate an image

Understanding Nodes

ComfyUI uses a node-based graph system. Key node types:

Node	Purpose
Load Checkpoint	Load a Stable Diffusion model
CLIP Text Encode	Convert text prompts to model-readable format
KSampler	The main generation node (controls steps, CFG, seed)
Empty Latent Image	Set the output image dimensions
VAE Decode	Convert latent space to a visible image
Save Image	Save the generated image to disk

Loading and Saving Workflows

bash

# Workflows are saved as JSON files
# Use the menu: Workflow > Save to save your current workflow

# Load a workflow from a JSON file
# Use the menu: Workflow > Open to load a saved workflow

# You can also drag and drop workflow JSON files onto the ComfyUI canvas

# Workflows are also embedded in generated PNG images
# Drag a ComfyUI-generated image onto the canvas to load its workflow

Installing ComfyUI Manager

ComfyUI Manager is an essential extension that adds a GUI for installing custom nodes, models, and workflows.

bash

# Navigate to the custom nodes directory
cd ComfyUI/custom_nodes

# Clone ComfyUI Manager
git clone https://github.com/ltdrdata/ComfyUI-Manager.git

# Restart ComfyUI

After restarting, a Manager button appears in the ComfyUI interface. From there you can:

Browse and install custom nodes
Search for missing nodes in a workflow
Update installed custom nodes
Install models directly from the UI

Popular Custom Nodes

Install these through ComfyUI Manager or manually:

Custom Node	Purpose
ComfyUI-Manager	Node and model management
ComfyUI-Impact-Pack	Advanced image processing nodes
ComfyUI-AnimateDiff	Video generation from images
ComfyUI-ControlNet-Auxiliary	ControlNet preprocessors
ComfyUI-KJNodes	Utility nodes for workflow building

Running as a Systemd Service

bash

# Create a systemd service file
sudo tee /etc/systemd/system/comfyui.service << 'EOF'
[Unit]
Description=ComfyUI Image Generation Server
After=network.target

[Service]
Type=simple
User=your-username
WorkingDirectory=/home/your-username/ComfyUI
ExecStart=/home/your-username/ComfyUI/venv/bin/python3 main.py --listen 0.0.0.0 --port 8188
Restart=on-failure
RestartSec=10

[Install]
WantedBy=multi-user.target
EOF

# Enable and start the service
sudo systemctl daemon-reload
sudo systemctl enable comfyui
sudo systemctl start comfyui

# Check the service status
sudo systemctl status comfyui

Update

Using comfy-cli

bash

# Update ComfyUI
comfy update

# Update all custom nodes
comfy nodes update all

Manual Update

bash

# Navigate to the ComfyUI directory
cd ComfyUI

# Pull the latest changes
git pull origin master

# Update dependencies
source venv/bin/activate
pip install -r requirements.txt

# Update custom nodes
cd custom_nodes
for dir in */; do (cd "$dir" && git pull 2>/dev/null); done

Troubleshooting

ComfyUI fails to start

bash

# Activate the virtual environment and check for errors
cd ComfyUI
source venv/bin/activate
python3 main.py

# If modules are missing, reinstall dependencies
pip install -r requirements.txt

# If the venv is corrupted, recreate it
rm -rf venv
python3 -m venv venv
source venv/bin/activate
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124
pip install -r requirements.txt

CUDA out of memory

bash

# Use low VRAM mode
python3 main.py --lowvram

# Reduce image resolution in the Empty Latent Image node
# Use 512x512 for SD 1.5 or 1024x1024 for SDXL

# Close other GPU-intensive applications

Workflow errors (red nodes)

Red nodes indicate missing custom nodes
Open ComfyUI Manager and click Install Missing Custom Nodes
Or manually install the missing node package from the custom_nodes directory

Generated images are black or corrupted

bash

# Try a different VAE
# Download and load the SDXL VAE explicitly in your workflow

# Check if the model file is corrupted
ls -lh models/checkpoints/

# Try running with --force-fp32 for precision issues
python3 main.py --force-fp32

Slow generation speed

bash

# Make sure CUDA is being used
python3 -c "import torch; print(torch.cuda.is_available())"

# Enable xformers for better performance
pip install xformers

# Use --highvram if you have plenty of VRAM (keeps models loaded)
python3 main.py --highvram

AI Tools Overview -- Overview of all AI tools on Debian
Stable Diffusion WebUI -- Traditional UI alternative
ComfyUI GitHub -- Official repository
ComfyUI Examples -- Official workflow examples
OpenArt Workflow Gallery -- Community-shared workflows

ComfyUI ​

Prerequisites ​

Installation ​

Method 1: comfy-cli (Recommended) ​

Method 2: Manual Installation ​

Verify PyTorch CUDA Setup ​

Download Models ​

Configuration ​

Launch Options ​

Model Directory Structure ​

Sharing Models with AUTOMATIC1111 ​

Usage ​

Basic Workflow ​

Understanding Nodes ​

Loading and Saving Workflows ​

Installing ComfyUI Manager ​

Popular Custom Nodes ​

Running as a Systemd Service ​

Update ​

Using comfy-cli ​

Manual Update ​

Troubleshooting ​

ComfyUI fails to start ​

CUDA out of memory ​

Workflow errors (red nodes) ​

Generated images are black or corrupted ​

Slow generation speed ​

Related Resources ​

ComfyUI

Prerequisites

Installation

Method 1: comfy-cli (Recommended)

Method 2: Manual Installation

Verify PyTorch CUDA Setup

Download Models

Configuration

Launch Options

Model Directory Structure

Sharing Models with AUTOMATIC1111

Usage

Basic Workflow

Understanding Nodes

Loading and Saving Workflows

Installing ComfyUI Manager

Popular Custom Nodes

Running as a Systemd Service

Update

Using comfy-cli

Manual Update

Troubleshooting

ComfyUI fails to start

CUDA out of memory

Workflow errors (red nodes)

Generated images are black or corrupted

Slow generation speed

Related Resources