Stable Diffusion WebUI (AUTOMATIC1111)

AUTOMATIC1111's Stable Diffusion WebUI is the most widely used web interface for AI image generation. It supports text-to-image, image-to-image, inpainting, outpainting, and dozens of advanced features through a browser-based UI. The project has a massive ecosystem of extensions, models, and community resources. It runs entirely on your local machine, giving you full control over the generation process.

Prerequisites

Before installing Stable Diffusion WebUI, make sure you have the following:

Debian 12 (Bookworm) or later -- 64-bit x86_64
Python 3.10 or later -- Required for the application and its dependencies
NVIDIA GPU with 4 GB+ VRAM -- 8 GB+ recommended for higher resolutions
20 GB+ free disk space -- For the application, models, and generated images
Git -- For cloning the repository

WARNING

While Stable Diffusion WebUI can run on CPU, it is extremely slow (minutes per image instead of seconds). An NVIDIA GPU is strongly recommended for a usable experience.

Installation

Step 1: Install System Dependencies

bash

# Update package lists
sudo apt update

# Install Python and required system packages
sudo apt install -y python3 python3-pip python3-venv python3-dev
sudo apt install -y git wget curl
sudo apt install -y libgl1-mesa-glx libglib2.0-0 libsm6 libxrender1 libxext6

Step 2: Install NVIDIA Drivers and CUDA

bash

# Install NVIDIA drivers (non-free repos must be enabled)
sudo apt install -y nvidia-driver firmware-misc-nonfree

# Reboot to load the driver
sudo reboot

# After reboot, verify the GPU
nvidia-smi

Step 3: Clone and Run the WebUI

bash

# Clone the repository
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
cd stable-diffusion-webui

# Launch the WebUI (first run will download dependencies and set up a venv)
./webui.sh

The first launch will take several minutes as it:

Creates a Python virtual environment
Downloads and installs PyTorch with CUDA support
Installs all Python dependencies
Downloads the default Stable Diffusion model

Once ready, open http://localhost:7860 in your browser.

Step 4: Download Additional Models

The default model is Stable Diffusion v1.5. You can download better models from Civitai or Hugging Face.

bash

# Navigate to the models directory
cd stable-diffusion-webui/models/Stable-diffusion

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

# Download the SDXL refiner (optional, for higher quality)
wget "https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/resolve/main/sd_xl_refiner_1.0.safetensors"

TIP

Civitai.com hosts thousands of community-created models, LoRAs, and embeddings. Download .safetensors files and place them in the appropriate models/ subdirectory.

Configuration

Launch Arguments

You can customize the WebUI behavior by editing the launch arguments.

bash

# Edit the webui-user.sh file to set persistent launch arguments
nano stable-diffusion-webui/webui-user.sh

Common launch arguments:

bash

# Example webui-user.sh configuration
export COMMANDLINE_ARGS="--xformers --enable-insecure-extension-access --listen"

Argument	Description
`--listen`	Listen on all network interfaces (access from other devices)
`--port 7860`	Set the web server port
`--xformers`	Enable xformers for faster, memory-efficient inference
`--medvram`	Reduce VRAM usage (for 4-6 GB VRAM GPUs)
`--lowvram`	Aggressively reduce VRAM usage (for 2-4 GB VRAM GPUs)
`--no-half`	Disable half-precision (fixes issues on some GPUs)
`--api`	Enable the REST API alongside the web UI
`--enable-insecure-extension-access`	Allow installing extensions from the UI
`--autolaunch`	Automatically open the browser on startup
`--theme dark`	Use dark theme

Low VRAM Options

If you have limited GPU memory, use these optimizations:

bash

# For 4-6 GB VRAM GPUs
./webui.sh --medvram --xformers

# For 2-4 GB VRAM GPUs
./webui.sh --lowvram --xformers

# For extremely limited VRAM (slow but functional)
./webui.sh --lowvram --no-half --opt-sub-quad-attention

Model Directory Structure

stable-diffusion-webui/models/
  Stable-diffusion/     # Main checkpoint models (.safetensors, .ckpt)
  Lora/                 # LoRA fine-tuned models
  VAE/                  # VAE models for improved image quality
  ControlNet/           # ControlNet models for guided generation
  ESRGAN/               # Upscaler models
  hypernetworks/        # Hypernetwork models
  embeddings/           # Textual inversion embeddings

Usage

Text-to-Image (txt2img)

Open http://localhost:7860 in your browser
Select a model from the Stable Diffusion checkpoint dropdown
Enter a text prompt describing the image you want
Optionally add a Negative prompt for things to avoid
Set the image dimensions, sampling steps, and CFG scale
Click Generate

Example prompts:

Prompt: a serene mountain landscape at sunset, photorealistic, 4k, detailed
Negative prompt: blurry, low quality, watermark, text, deformed

Image-to-Image (img2img)

Click the img2img tab
Upload a source image
Enter a prompt describing the desired modifications
Adjust the Denoising strength (0.3-0.5 for subtle changes, 0.7-0.9 for major changes)
Click Generate

Inpainting

Click the img2img tab, then select Inpaint
Upload an image and paint over the area you want to change
Enter a prompt describing what should replace the painted area
Click Generate

ControlNet

ControlNet allows precise control over image generation using reference images (poses, depth maps, edges).

bash

# Install the ControlNet extension
cd stable-diffusion-webui/extensions
git clone https://github.com/Mikubill/sd-webui-controlnet.git

# Download ControlNet models
cd ../models/ControlNet
wget "https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_openpose.pth"

After restarting the WebUI, a ControlNet section appears in the txt2img and img2img tabs.

LoRA (Low-Rank Adaptation)

LoRA models are small fine-tuned layers that modify the base model's output style or subject.

bash

# Place LoRA files in the Lora directory
cp my-lora-model.safetensors stable-diffusion-webui/models/Lora/

In the prompt, reference a LoRA with the syntax: <lora:model-name:weight>, for example:

a portrait photo <lora:my-style:0.8>, detailed, realistic

API Usage

Enable the API with --api and interact programmatically.

bash

# Start with API enabled
./webui.sh --api

# Generate an image via API
curl -X POST http://localhost:7860/sdapi/v1/txt2img \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "a beautiful sunset over the ocean",
    "negative_prompt": "blurry, low quality",
    "steps": 25,
    "width": 512,
    "height": 512,
    "cfg_scale": 7
  }'

Running as a Systemd Service

bash

# Create a systemd service file
sudo tee /etc/systemd/system/stable-diffusion.service << 'EOF'
[Unit]
Description=Stable Diffusion WebUI
After=network.target

[Service]
Type=simple
User=your-username
WorkingDirectory=/home/your-username/stable-diffusion-webui
ExecStart=/home/your-username/stable-diffusion-webui/webui.sh --listen --xformers --api
Restart=on-failure
RestartSec=10

[Install]
WantedBy=multi-user.target
EOF

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

Update

bash

# Navigate to the WebUI directory
cd stable-diffusion-webui

# Pull the latest changes
git pull origin master

# Restart the WebUI (dependencies will be updated automatically)
./webui.sh

Troubleshooting

"CUDA out of memory" error

bash

# Use lower VRAM mode
./webui.sh --medvram --xformers

# Reduce image resolution (use 512x512 instead of 1024x1024)
# Reduce batch size to 1
# Close other GPU-intensive applications

# For extreme cases
./webui.sh --lowvram --no-half --opt-sub-quad-attention

WebUI fails to start

bash

# Check for Python version
python3 --version

# Remove the virtual environment and let it recreate
rm -rf stable-diffusion-webui/venv

# Restart the WebUI
./webui.sh

Black or corrupted images

bash

# Try disabling half-precision
./webui.sh --no-half

# Or use a specific VAE model
# Download a VAE and place it in models/VAE/

# Check GPU driver compatibility
nvidia-smi

Extensions not loading

bash

# Make sure extension access is enabled
./webui.sh --enable-insecure-extension-access

# Update all extensions
cd stable-diffusion-webui/extensions
for dir in */; do (cd "$dir" && git pull); done

Slow generation speed

bash

# Enable xformers for better performance
./webui.sh --xformers

# Make sure you are using the GPU (not CPU)
# Check nvidia-smi while generating -- GPU utilization should be high

# Use smaller image sizes for faster generation
# Use fewer sampling steps (20-25 is usually sufficient)

AI Tools Overview -- Overview of all AI tools on Debian
ComfyUI -- Node-based alternative with more flexibility
AUTOMATIC1111 Wiki -- Official documentation
Civitai -- Community models, LoRAs, and resources
Hugging Face -- Model hosting and downloads

Stable Diffusion WebUI (AUTOMATIC1111) ​

Prerequisites ​

Installation ​

Step 1: Install System Dependencies ​

Step 2: Install NVIDIA Drivers and CUDA ​

Step 3: Clone and Run the WebUI ​

Step 4: Download Additional Models ​

Configuration ​

Launch Arguments ​

Low VRAM Options ​

Model Directory Structure ​

Usage ​

Text-to-Image (txt2img) ​

Image-to-Image (img2img) ​

Inpainting ​

ControlNet ​

LoRA (Low-Rank Adaptation) ​

API Usage ​

Running as a Systemd Service ​

Update ​

Troubleshooting ​

"CUDA out of memory" error ​

WebUI fails to start ​

Black or corrupted images ​

Extensions not loading ​

Slow generation speed ​

Related Resources ​

Stable Diffusion WebUI (AUTOMATIC1111)

Prerequisites

Installation

Step 1: Install System Dependencies

Step 2: Install NVIDIA Drivers and CUDA

Step 3: Clone and Run the WebUI

Step 4: Download Additional Models

Configuration

Launch Arguments

Low VRAM Options

Model Directory Structure

Usage

Text-to-Image (txt2img)

Image-to-Image (img2img)

Inpainting

ControlNet

LoRA (Low-Rank Adaptation)

API Usage

Running as a Systemd Service

Update

Troubleshooting

"CUDA out of memory" error

WebUI fails to start

Black or corrupted images

Extensions not loading

Slow generation speed

Related Resources