● LIVEReading: Run Stable Diffusion Locally With a Full Web UITotal time: 12 minSteps: 5Worked first time: 55%● LIVEReading: Run Stable Diffusion Locally With a Full Web UITotal time: 12 minSteps: 5Worked first time: 55%
CBW
Spicygithub.com/AUTOMATIC1111/stable-diffusion-webui2026-05-18

Run Stable Diffusion Locally With a Full Web UI

Install AUTOMATIC1111's Stable Diffusion web UI on your own machine and generate images from text prompts — no cloud account needed. Works on Windows and Linux with an Nvidia GPU.

// Build stats

  • Total time12 min
  • Number of steps5
  • DifficultySpicy
  • Worked first time55%
// Before you start

What you need

  • Windows 10/11 or Linux (Mac is possible but slower)
  • Nvidia GPU with at least 4 GB VRAM (8 GB recommended)
  • Python 3.10.6 installed and added to PATH (Windows)
  • Git installed
  • 10–20 GB free disk space for models and outputs
  • Stable internet connection for first-time downloads
01
Step 1 of 5

Install Python 3.10.6 and Git

5 min

The web UI needs a specific Python version — 3.10.6. Newer versions break the AI libraries it depends on. During the Python installer, tick the box that says 'Add Python to PATH' before clicking Install. Git is used to download the project files. If you are on Linux, your package manager handles both.

Terminal · mac
$ Windows: download Python 3.10.6 from https://www.python.org/downloads/release/python-3106/ and Git from https://git-scm.com/download/win — run both installers.
$
$ Linux (Debian/Ubuntu):
$ sudo apt install wget git python3 python3-venv libgl1 libglib2.0-0
What you should see
Python installer finishes with no errors. Running 'python --version' in a terminal shows Python 3.10.6.
This might happen

Python 3.11 or 3.12 is already installed and torch fails to install later.

Uninstall the newer Python version, install 3.10.6, and make sure it appears first in your PATH.

02
Step 2 of 5

Clone the repository

2 min

This command downloads all the project files from GitHub into a folder called 'stable-diffusion-webui' on your computer. Open a terminal (Command Prompt on Windows, or a terminal app on Linux) and run the command below. Navigate to wherever you want the folder to live first — your Desktop or Documents folder works fine.

Terminal · mac
$ git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
What you should see
A folder named 'stable-diffusion-webui' appears in your current directory containing many files.
This might happen

'git' is not recognized as a command.

Git was not added to PATH during installation. Reinstall Git and check the option to add it to PATH, then open a fresh terminal.

03
Step 3 of 5

Download a Stable Diffusion model checkpoint

10 min

The web UI is just the interface — it needs an actual AI model file to generate images. The most common starting point is a .safetensors file from Hugging Face or Civitai. Download one and place it in the correct folder. Without this file the UI will launch but cannot generate anything.

Terminal · mac
$ Place your downloaded .safetensors or .ckpt model file into this folder:
$
$ Windows: stable-diffusion-webui\models\Stable-diffusion\
$ Linux: stable-diffusion-webui/models/Stable-diffusion/
$
$ Example free model to download: https://huggingface.co/runwayml/stable-diffusion-v1-5
What you should see
The models/Stable-diffusion folder contains at least one .safetensors or .ckpt file.
This might happen

You are not sure which model to pick.

Start with 'v1-5-pruned-emaonly.safetensors' from the Runway ML page on Hugging Face — it is small (~4 GB) and widely supported.

04
Step 4 of 5

Launch the web UI for the first time

10–20 min (first run only)

On first launch the startup script automatically creates a Python virtual environment, downloads PyTorch, and installs all required libraries. This can take 10–20 minutes and uses several gigabytes of data. Subsequent launches take under a minute. On Windows double-click the .bat file; on Linux run the shell script from your terminal inside the project folder.

Terminal · mac
$ Windows: double-click webui-user.bat
$
$ Linux:
$ cd stable-diffusion-webui
$ bash webui.sh
What you should see
The terminal prints many download and install lines, then ends with a line like: 'Running on local URL: http://127.0.0.1:7860'
This might happen

Installation stalls or throws a CUDA / torch error.

Make sure your Nvidia drivers are up to date. Visit https://developer.nvidia.com/cuda-downloads and install the latest driver for your GPU, then rerun the launch script.

05
Step 5 of 5

Open the UI and generate your first image

2 min

Once the terminal shows the local URL, open your web browser and go to that address. You will see the full Stable Diffusion interface. Type a description of what you want in the top text box (the 'prompt'), leave all other settings at their defaults, and click the orange Generate button. Your first image will appear on the right side of the screen.

Terminal · mac
$ Open your browser and go to:
$ http://127.0.0.1:7860
$
$ Type a prompt such as:
$ a photograph of a red fox sitting in a snowy forest, golden hour lighting
$
$ Click Generate.
What you should see
An image matching your description appears on the right side of the page within 10–60 seconds depending on your GPU.
This might happen

Generation is extremely slow or the page shows an out-of-memory error.

Add --medvram to the COMMANDLINE_ARGS line inside webui-user.bat (Windows) or webui.sh (Linux), save the file, and relaunch. For very low VRAM cards use --lowvram instead.

// Status

cooked. baked. worked.

A locally running web interface at http://127.0.0.1:7860 where you can type text prompts and generate AI images entirely on your own machine, with no usage fees or cloud accounts.

// the honest bit

The honest part

This project has a lot of moving parts — Python version sensitivity, GPU drivers, and large model downloads all create friction. Non-developers often hit a snag on the first run. The automatic installer handles most dependencies, but CUDA driver mismatches and wrong Python versions are the two most common blockers. AMD GPU support exists but is significantly harder to set up and is not covered here. Mac support via MPS is available but noticeably slower than Nvidia. Model files are large (2–7 GB each) and must be sourced separately — the repo ships with no model included.