Install Git for Windows.

Github is a repository, or storage place for Open Source software. Developers from around the world host their latest code with Github. Don’t worry, its very safe to follow all of these instructions. This app will allow all programs hosted on Github to update.

https://gitforwindows.org/

Click “Next” to all defaults. YES you want Git in the PATH environment.

Step 2:

Installing Python. This is a programming language. Having it on your Windows system will allow your new Github download to “compile” into a full program.

https://www.python.org/downloads/release/python-3106/

Make sure to check “Add Python 3.10 to PATH" This is very important, and do not install the latest version You must use 3.10.6

Requirements:

Your PC should run Windows 10 or higher with a Nivdia GPU with 4 GB VRAM or more. Just like GPU rendering, the higher the resolution image you generate, the more VRAM will be consumed.

All the software required is free and open source. We are going to use Stable Diffusion as the backbone and to interface with Automatic1111. If you’ve ever experimented with open source research software from Github, you’ll know that in the early days of development the only way to process images is through text commands. Automatic1111 replaces the need to text commands to generate images.

Step 3:

It may be 2024, but we need to go back to DOS and open a Command Prompt Window.

Search and type: CMD

A black screen will appear. Type:

cd C:\

or

cd ..

cd ..

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

It should look like this in the image below. cd .. (means change directory back 1)

Step 4:

You can now close Command Prompt and return to Windows File Explorer. In the future, if you want to try new programs from Github. “git clone” is often the first step.

We need to edit the “webui-user.bat” file using notepad. This is the file we use every time we want to run Stable Diffusion. But first we need to edit the defaults.

Step 5:

We can now run the webui-user.bat file, and the command window will return and start to download and install dependencies automatically. Once complete it will be ready to go, but first we need to update it further.

“—api” allows access to the Photoshop plugin and other extensions

“—xformers” will increase speed

“—medvram” or “lowvram” can be added to the command line if you have less than 12GB or 8GB

“git pull” will update SD everytime you run.

Step 6:

The model that comes with SD isn’t very good, so we need to download a more recently trained model for our needs. A “model” also called a “checkpoint” is a single file that contains ALL of the imagery we use when generating AI art. On first run you may have notice a 4GB file that was downloaded, that is Stable Diffusion v1.5, which is in our model directory here:

C:\stable-diffusion-webui\models\Stable-diffusion

https://civitai.com/models/132632

Version 1.5 models contain mages that were gathered from the internet at 512 x 512 resolution. SD 2.0 used 768x768 but was never popular and never caught on for some reason. The latest is SDXL which is native 1024 x 1024 and more cinematic, meaning it has better initial composition. I still use SD 1.5 for editing my renderings, and this is my favourite set from CivitAI. The repository for all things Stable Diffusion.

Final Step: READY TO GO

Finally you can launch Stable Diffusion.

Run the webui-user.bat file, and the command window will return and display a URL. Copy that that into your browser if it does not automatically appear.

If you already have Automatic1111 running, click the refresh blue circle arrow button at the top left corner. You should now have 2 or more checkpoint models to choose from.