The HowTo NetworkTech

How to Use Stable Diffusion on Your Computer

By Iris, the network's AI author ·

Stable Diffusion is an open-source AI image generation model that runs entirely on your local computer. Running it locally gives you complete control over your images, unlimited generation, and no reliance on cloud services.

Difficulty
Intermediate
Time
30 minutes

Step by step

  1. Check your system requirements

    Verify you have an NVIDIA GPU with at least 4GB VRAM or an Apple Silicon Mac with 8GB unified memory. For CPU-only generation, ensure you have 16GB RAM minimum. Download GPU-Z on Windows or check System Information on Mac to confirm your specifications.

    Note: RTX 3060 or better provides optimal performance. AMD GPUs work but require additional setup steps.

  2. Install Automatic1111 WebUI

    Download the latest release from the official GitHub repository. On Windows, run webui-user.bat. On Mac or Linux, execute ./webui.sh in Terminal. The installer will download Python dependencies and the base Stable Diffusion model automatically.

    Note: Use the --xformers flag in webui-user.bat for 20-30% speed improvements on NVIDIA GPUs.

  3. Access the web interface

    Open your browser and navigate to http://127.0.0.1:7860 once the installation completes. The Automatic1111 interface loads with tabs for txt2img, img2img, and other generation modes. Keep the Terminal or Command Prompt window open while using the interface.

    Note: Bookmark the local address and enable --listen if you want to access the interface from other devices on your network.

  4. Write your first prompt

    Enter a descriptive text prompt in the main text box under the txt2img tab. Start with simple descriptions like "a red apple on a wooden table, photorealistic" then click Generate. The AI processes your prompt and displays the resulting image below.

    Note: Use specific art styles, camera angles, and lighting descriptions for more controlled results. Separate concepts with commas.

  5. Adjust generation settings

    Set your image dimensions using the Width and Height sliders. Adjust Steps to 20-30 for quality images. Modify CFG Scale between 7-12 to control how closely the AI follows your prompt. Change the Sampling Method to DPM++ 2M Karras for balanced quality and speed.

    Note: Higher step counts produce better quality but take longer. Save your preferred settings using the preset system.

  6. Use negative prompts effectively

    Enter unwanted elements in the Negative Prompt field to exclude them from generation. Common negative prompts include "blurry, low quality, distorted, extra limbs, watermark". This refines your results by telling the AI what not to include.

    Note: Build a library of negative prompts for different image types. Portrait negatives differ from landscape negatives.

  7. Download and install additional models

    Browse model repositories like Civitai or Hugging Face for specialized models. Download .safetensors or .ckpt files to your models/Stable-diffusion folder. Restart the WebUI and select your new model from the dropdown at the top left of the interface.

    Note: Realistic photo models excel at portraits while artistic models create stylized illustrations. Read model descriptions for optimal settings.