What you need before you start
Stable Diffusion is free software that runs on your own computer and generates images from text descriptions. Unlike web-based tools, it uses your graphics card (GPU) to do the work, which is why installation matters — the process differs depending on whether you have an Nvidia card, an AMD card, or neither. You will need at least 6 GB of video memory to run it smoothly, though 8 GB or more is better.
Before you install anything, check what graphics card you have. On Windows, right-click your desktop, look for "Nvidia Control Panel" or "AMD Radeon Settings" in the menu, or open Device Manager and expand "Display adapters" to see the name. On Mac, click the Apple menu, select "About This Mac", and look under "Graphics". If you see Intel integrated graphics only, Stable Diffusion will run much slower, but it will still work.
You will also need about 25 GB of free disk space for the software and the model files it downloads. Have that space ready before you begin.
Key Takeaways
- Stable Diffusion runs locally on your computer using your graphics card, and installation steps differ for Nvidia, AMD, and Intel setups.
- The easiest installation path for most people is using a launcher like Automatic1111 or Comfy UI, which handles dependencies automatically instead of requiring command-line work.
- Your graphics card memory (6 GB minimum, 8 GB recommended) determines how fast images generate and how large they can be.
- After installation, you read the actual model file separately — this is the AI weights that do the image generation, not included in the software itself.
- The first run downloads additional files automatically, which can take 10 to 30 minutes depending on your internet speed.
Installing on Windows with an Nvidia card
Nvidia cards are the easiest to set up because Stable Diffusion was developed and tested primarily on Nvidia hardware. read Automatic1111, which is a launcher that handles all the technical setup for you. Go to github.com/AUTOMATIC1111/stable-diffusion-webui, scroll down to the "Installation and Running" section, and read the file labeled "stable-diffusion-webui-master.zip" (or the latest version available). Extract this folder to a location you will remember — your Documents folder or Desktop works fine.
Open the extracted folder and double-click the file named "webui-user.bat". A command window will open and begin downloading files. This first run takes 10 to 30 minutes because it is downloading PyTorch (the AI framework), the Stable Diffusion code, and model files. Do not close the window. When it finishes, you will see a message with a web address like "http://127.0.0.1:7860" — copy this into your web browser and you will see the Stable Diffusion interface.
Before you can generate images, you need the actual model file. The launcher comes with a default model, but you can read others from Hugging Face (huggingface.co/models?pipeline_tag=text-to-image). read a model file (usually named something like "sd-v1-5.safetensors") and place it in the "models/Stable-diffusion" folder inside your extracted Stable Diffusion folder. Restart the launcher and the new model will appear in the dropdown menu.
Installing on Windows with an AMD card
AMD cards require an extra step because they use a different framework called ROCm. read the same Automatic1111 launcher from github.com/AUTOMATIC1111/stable-diffusion-webui. Before you run it, you need to install AMD's HIP drivers, which let your AMD card communicate with the software. Go to rocmdocs.amd.com/en/docs/deploy/linux/index.html, find the Windows section, and follow the installation steps for your specific card model.
After HIP is installed, extract the Stable Diffusion folder and open "webui-user.bat" as before. The first run will take longer than Nvidia because it is compiling code for your specific card, but the process is the same. When you see the web address, open it in your browser. AMD cards often run 20 to 40 percent slower than equivalent Nvidia cards, but the results are identical.
read a model file from Hugging Face and place it in the "models/Stable-diffusion" folder, then restart the launcher. Some AMD users report better performance with specific model versions — if generation is slow, try a smaller model like "sd-v1-5-pruned" instead of the full version.
Installing on Mac
Mac installation is simpler because you do not need to install drivers, but performance is slower because Macs use integrated graphics instead of dedicated cards. read Automatic1111 from github.com/AUTOMATIC1111/stable-diffusion-webui and extract the folder. Open Terminal (find it in Applications > Utilities), type "cd " (with a space after it), then drag the extracted folder into the Terminal window and press Enter. This navigates Terminal to that folder.
Type "bash webui.sh" and press Enter. The first run downloads files and takes 15 to 45 minutes depending on your internet speed and Mac model. When you see the web address, open it in your browser. On Mac, generation is slower — a single image might take 2 to 5 minutes instead of 30 seconds on a good Nvidia card — but it works.
read a model file from Hugging Face and place it in the "models/Stable-diffusion" folder. For Mac, smaller models like "sd-v1-5-pruned" or "sd-v2-1-base" run faster than full models. Restart the launcher and select your model from the dropdown.
Downloading and organizing model files
The model file is the actual AI that generates images — the launcher is just the interface. Hugging Face (huggingface.co) hosts most public models. Search for "Stable Diffusion" and look for files ending in ".safetensors" or ".ckpt". Click the read button and save the file to your computer.
Move the downloaded file into the "models/Stable-diffusion" folder inside your Stable Diffusion installation. If that folder does not exist, create it. The full path should look like "C:\Users\YourName\Documents\stable-diffusion-webui\models\Stable-diffusion" on Windows or "~/stable-diffusion-webui/models/Stable-diffusion" on Mac. Restart the launcher and your new model will appear in the dropdown menu at the top of the web interface.
Different models produce different styles. "Stable Diffusion v1.5" is general-purpose and fast. "Deliberate" is better for realistic images. "DreamShaper" is better for stylized art. You can have multiple models in the folder and switch between them without restarting.
Running Stable Diffusion after installation
After the first installation, you do not need to repeat the setup. On Windows, double-click "webui-user.bat" in your Stable Diffusion folder. On Mac, open Terminal, navigate to the folder (type "cd " and drag the folder in), and type "bash webui.sh". The launcher will start in 30 to 60 seconds and display the web address.
Open that address in your browser. You will see a text box where you type what you want to generate, settings for image size and quality, and a "Generate" button. Type a description like "a red cat sitting on a blue chair" and click Generate. The first image takes longer because the software is loading the model into your graphics card memory. Subsequent images are faster.
If generation is very slow or crashes, you may not have enough graphics card memory. Lower the image resolution (512x512 instead of 768x768), enable "Attention slicing" in the settings, or try a smaller model file. These changes make generation slower but use less memory.
Troubleshooting common problems
If the launcher does not start, check that you extracted the folder correctly and that the path does not contain spaces or special characters. Move the folder to a simpler location like "C:\Stable-Diffusion" on Windows. If you see an error about CUDA or ROCm, your graphics drivers are out of date — update them through Nvidia's or AMD's official website, not through Windows Update.
If generation crashes or says "out of memory", your graphics card does not have enough space for the model and the image at the same time. Lower the resolution, enable "Memory efficient attention" in the settings, or use a smaller model. On Mac, this is common — try "sd-v1-5-pruned" instead of the full model.
If the web interface does not load, check that no other program is using port 7860. Close the launcher, wait 10 seconds, and restart it. If you still see an error, try typing "http://localhost:7860" instead of the address shown in the Terminal window.
Frequently Asked Questions
Do I need an internet connection to run Stable Diffusion?
You need internet for the initial read and installation, which can take 30 minutes to an hour. After that, Stable Diffusion runs entirely on your computer — you can generate images offline. You only need internet again if you want to read new model files.
Will Stable Diffusion slow down my computer when it is not running?
No. Stable Diffusion only uses your graphics card and disk space when you actually run the launcher and generate images. When the launcher is closed, it uses nothing. Your computer runs at normal speed.
Can I use Stable Diffusion on a laptop?
Yes, but performance depends on your graphics card. Laptops with Nvidia RTX cards work well. Laptops with AMD or Intel integrated graphics work but are slower — expect 2 to 5 minutes per image instead of 30 seconds. Make sure you have at least 25 GB of free disk space.
What is the difference between Automatic1111 and other launchers?
Automatic1111 is the most popular and has the most features and community support. Comfy UI is another option that is more visual and easier for beginners. Both do the same thing — they run Stable Diffusion on your computer. Choose whichever interface you prefer.
Do I need to pay for Stable Diffusion or the models?
No. Stable Diffusion, Automatic1111, and most public models are free. You only pay for the electricity to run your computer. Some specialized models on Hugging Face may have restrictions, but the main ones are free to read and use.