What you need before you start
Stable Diffusion is software that runs on your own computer to generate images from text descriptions. Unlike web-based tools, you download and install it locally, which means you control it entirely and don't send your prompts to anyone else's servers. The trade-off is that your computer needs enough power to handle it.
Before you begin, check that your system meets the minimum requirements. You need at least 8 GB of RAM, though 16 GB or more runs noticeably faster. Your graphics card matters most: an NVIDIA card with at least 4 GB of VRAM (video memory) is ideal, but AMD and Intel GPUs work too. If you have an older or basic laptop without a dedicated graphics card, Stable Diffusion will still run but will be very slow—generating a single image might take several minutes instead of seconds.
You'll also need about 25 GB of free disk space for the software and the model files it downloads. Have a text editor ready (Notepad on Windows or TextEdit on Mac work fine) because you'll need to enter a few commands, but nothing complicated.
Key Takeaways
- Stable Diffusion requires Python, which you install first, then a launcher application that handles the rest of the setup automatically.
- The easiest route for most people is AUTOMATIC1111's web UI, which runs in your browser and doesn't require typing commands after the initial setup.
- Your graphics card type (NVIDIA, AMD, or Intel) determines which version you download, so check your system settings before you start.
- The first launch downloads large model files automatically, which takes 10 to 30 minutes depending on your internet speed.
- Once installed, you run Stable Diffusion by opening a single file each time, and it opens in your web browser automatically.
Install Python first
Stable Diffusion runs on Python, a programming language your computer needs to have installed. Go to python.org, click the Downloads button, and download Python 3.10 or 3.11 (not the newest version—Stable Diffusion hasn't caught up yet). Run the installer.
On Windows, the installer shows a checkbox that says "Add Python to PATH"—check that box before you click Install. This step matters; if you skip it, your computer won't find Python later. On Mac, just run through the installer normally. On Linux, Python is usually already there, so skip this step.
After installation finishes, open a command prompt (Windows) or terminal (Mac/Linux) and type python --version. You should see the version number you just installed. If you see an error instead, Python didn't install correctly—go back and run the installer again, making sure to check the PATH box on Windows.
Download and run AUTOMATIC1111's web UI
AUTOMATIC1111 is a launcher that handles all the complicated setup for you. Go to github.com/AUTOMATIC1111/stable-diffusion-webui and look for the green "Code" button. Click it and select "Download ZIP". Save the file and extract it to a folder on your computer—your Desktop or Documents folder works fine. Name the folder something simple like "stable-diffusion".
Inside that folder, you'll see several files. On Windows, double-click the file called webui-user.bat. On Mac or Linux, open a terminal, navigate to that folder, and type bash webui-user.sh. A command window will open and start downloading files. This first run takes 10 to 30 minutes because it's downloading the model files (the actual AI that generates images). Your internet speed matters here—if you have slow internet, go get coffee.
When the download finishes, you'll see a message that says something like "Running on http://127.0.0.1:7860". Open your web browser and go to that address. You should see a page with a text box where you can type image descriptions. If you see an error instead, wait another minute—sometimes it takes a moment to fully start up.
Choose the right version for your graphics card
Before you run the launcher, you may need to tell it which type of graphics card you have. Open the folder where you extracted Stable Diffusion and find the file called webui-user.bat (Windows) or webui-user.sh (Mac/Linux). Open it with a text editor.
Look for a line that starts with set COMMANDLINE_ARGS= (Windows) or export COMMANDLINE_ARGS= (Mac/Linux). If you have an NVIDIA card, add --xformers at the end of that line. If you have an AMD card, add --opt-split-attention. If you have an Intel Arc GPU, add --use-intel-ext-attention. If you don't have a dedicated graphics card at all, add --cpu, but understand that image generation will be slow.
Save the file and close the text editor. The next time you run the launcher, it will use the settings you just entered. If you're not sure what graphics card you have, check your system settings: on Windows, right-click the desktop and look for "NVIDIA Control Panel" or "AMD Radeon Settings". On Mac, click the Apple menu, select "About This Mac", and look under "Graphics".
Generate your first image
Once the web page loads, you'll see a text box labeled "Prompt" on the left side. Type a description of what you want to see—something like "a cat wearing sunglasses, oil painting" or "a wooden cabin in the mountains at sunset". Be specific; vague descriptions produce vague results.
Below the prompt box, you'll see settings like "Steps" (usually 20 to 30) and "Sampling method" (usually "Euler" or "DPM++"). Leave these at their defaults for now. Click the blue "Generate" button and wait. On a fast computer with a good graphics card, this takes 5 to 15 seconds. On a slower system, it might take a minute or two.
The image appears on the right side of the screen. If you don't like it, change your prompt and click Generate again. Each image costs nothing—you're running this on your own computer, not paying a service. Experiment with different descriptions to see what works.
Troubleshooting common problems
If the web page won't load, check that the command window is still running in the background. If it closed, run the launcher again. If you see an error about "CUDA" or "GPU memory", your graphics card doesn't have enough memory. Go back to the webui-user file and add --medvram or --lowvram to the COMMANDLINE_ARGS line to use less memory at the cost of speed.
If image generation is extremely slow (more than a minute per image), your computer may not have a graphics card, or the launcher didn't detect it correctly. Check the command window for error messages. If you see "CPU" mentioned, the launcher is running on your processor instead of your graphics card. Close everything, check that you added the right graphics card flag to the webui-user file, and try again.
If you run out of disk space during the first launch, the download will fail. You need at least 25 GB free. Delete some files to make room, then run the launcher again—it will resume the download from where it stopped.
Running Stable Diffusion after the first time
You don't need to reinstall anything. Each time you want to use Stable Diffusion, just run the webui-user file again (double-click on Windows, or run the bash command on Mac/Linux). The command window opens, and after a few seconds, your browser loads the page automatically. Close the command window when you're done.
If you want to update Stable Diffusion to a newer version, go back to the GitHub page, download the latest ZIP file, and extract it to a new folder. Your old installation stays where it is. Many people keep multiple versions installed so they can switch between them.
Frequently Asked Questions
Can I use Stable Diffusion on a Mac with Apple Silicon?
Yes. Apple Silicon Macs (M1, M2, M3) run Stable Diffusion well. When you run the launcher, it detects your Mac automatically and uses the right settings. Image generation is faster than on Intel Macs but slower than on a high-end NVIDIA card.
What if I don't want to type commands in a terminal?
AUTOMATIC1111's web UI is designed so you don't have to type commands after the initial setup. You only type one command to start it (or double-click a file on Windows), then everything else happens in your browser. If even that feels like too much, other launchers like Invoke AI or ComfyUI have graphical installers that avoid the command line entirely.
Do I need an internet connection after installation?
You need internet for the first launch to download the model files. After that, Stable Diffusion runs entirely offline. You can generate images without being connected to the internet.
How much disk space do different model sizes need?
The standard Stable Diffusion 2 model is about 5 GB. The full installation with the launcher and dependencies takes about 25 GB total. Larger models like Stable Diffusion XL need an additional 7 GB. You can install multiple models and switch between them, but each one takes disk space.
Can I use Stable Diffusion commercially?
The software itself is open source and free to use. The images you generate are yours to use as you see fit, including commercially. However, check the license terms for any additional models you download, as some have restrictions on commercial use.