--A MORON'S GUIDE TO VOLD'S STABLE DIFFUSION GUI--

So you want to take the dive into Stable Diffusion and the wide world of AI generated art, but you don’t know where to start. This is a guide to set up what I believe to be the best UI based distribution currently available. It’s frequently updated with features from various distributions and it’s free of the Patreon begging, Discord integrating, and potential malware bull shit that plagues many mainstream distributions. Going forward, I’m going to refer to this version as Voldemort’s, or Voldy’s for short. You can inspect the code for Voldy’s here. Stable Diffusion only works well with a modern NVidia graphics card. If you’re running AMD or an old card, you’re probably shit out of luck.

• Python
• Git

Step 1: Download the model checkpoint (currently 1.4) from huggingface. You have to accept the license by registering for huggingface, which can be done by clicking a button that says Access repository and creating an account. There are two model checkpoint weights you can download, I’d personally recommend using sd-v1-4-full-ema.ckpt. It’s double the size so the download will take a bit, but it can lead to better results. It's apparently a bit of a memory hog though, so if you're running into memory issues, you may want to use sd-v1-4.ckpt instead.

Step 2: Once the download is complete, rename your chosen .ckpt file to model.ckpt.

Step 3: Download GPFGAN, which fixes the borked faces Stable Diffusion produces most of the time. CodeFormer now comes standard with Voldy's, which does the same thing and produces better results for the most part, but this is still nice to experiment around with.

Step 4: Download ESRGAN models for upscaling final results.

• Remacri is good for landscapes and structures.
• Lollypop is better for upscaling anything that has anthropomorphic figures.
• If you want to try out different models, you can find them here. Anything that ends with .pth can be loaded by Voldy’s.

Step 5: In file explorer, create (or choose) a root directory (location on your PC where Voldy's will be stored) to download the files to. In my case, I cloned the files directly to my C:\ drive, leading to a root directory location of: C:\

Step 6: Open command prompt in this directory. To open command prompt, press the Windows key, type cmd, and hit Enter. In the window that opens, enter the following command (you can copy and paste this and all following commands):

cd C:\

Alternatively, you can open command prompt to this directory through file explorer. To do so, make sure file explorer is in your newly created (or selected) directory and click on the navigation bar (which should say something along the lines of This PC > OS Disk (C:)). Clear anything on this bar, then type cmd and hit enter. Command prompt should open in the proper directory folder.

Step 7: To confirm you're in the right spot, enter the following command, which should send back your directory location. In my case it sent: `C:\”.

echo %cd

Step 8: Enter the following command, which will create a folder with the necessary files.

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

Step 9: In file explorer, navigate to the newly created stable-diffusion-webui directory (to confirm you're in the right place, look for a file called webui.bat). Move the model.ckpt and GFPGANv1.3.pth files you downloaded earlier into this directory.

Step 10: In the same directory, right click on webui-user.bat, click edit and replace its contents with the following, saving your changes.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--medvram --opt-split-attention

git pull
start http://127.0.0.1:7860/

call webui.bat

Step 11: Navigate to ESRGAN. Move the .pth ESRGAN files you downloaded earlier, such as Remacri or Lollipop, into this folder.

Step 12: Navigate back to your root directory and double click webui-user.bat to launch. It will download some files the first time, so be patient. Once Voldy’s is fully launched, it should return a URL of http://127.0.0.1:7860/ within the command prompt window. The launcher should have already navigated you there (you may need to refresh the page), but if accidentally closed it out and need to relaunch, go to that URL to view the UI and begin to render. To exit Voldy’s, close the command prompt window.

Once you launch Voldy’s, I would recommend setting your output folders (unless you’d prefer they output within your root directory). I was going to go hands off for this portion, but I'm feeling benevolent so I created a .bat to make this easier.

Step 1: Download the file below, and move it to wherever you want your images to output to. Run it, and it will create the appropriate sub-folders. You can delete the .bat after this is complete.

• Hand holding directory assignment .bat for friends

Step 2: Go to the Settings tab of the UI and assign your new file locations accordingly. Once you’ve assigned the locations, make sure to hit Apply Settings at the top of the page.

I’d also recommend enabling the following option, which will create subfolders for each new prompt you try. There are other options available that I go through in my guide on how to use Voldy’s.

• When writing images/grids, create a directory with name derived from the prompt

If it's not automatically selected, you should also select CodeFormer as your face restoration model. I personally set my weight to 0.5. As with the above, and anytime you make setting on the Settings tab, make sure you hit Apply Settings at the top of the page.

To run online: set COMMANDLINE_ARGS=--medvram --opt-split-attention --share

• You will see an xxx.app.gradio URL, which can be used to access from your preferred device.
• To assign a specific port, use the following command, replacing xxxx with your specified port.
  • set COMMANDLINE_ARGS=--medvram --opt-split-attention --share --port xxxx
  • Remember that all ports below 1024 need root/admin rights, for this reason it is advised to use a port above 1024.
• You may set up authentication for said gradio shared instance with the following command line, replacing user and pw with your selections:
  • set COMMANDLINE_ARGS=--medvram --opt-split-attention --share --gradio-auth username:password
  • Optionally, you can provide multiple sets of usernames and passwords separated by commas (user1:pw1, user2:pw2).

To run locally: set COMMANDLINE_ARGS=--medvram --opt-split-attention --listen

• This will allow computers on the same local network as your device to access the UI by going to the URL http://127.0.0.1:7860/.
• To assign a specific port, use the following command, replacing xxxx with your specified port.
  • set COMMANDLINE_ARGS=--medvram --opt-split-attention --listen --port xxxx
  • Remember that all ports below 1024 need root/admin rights, for this reason it is advised to use a port above 1024.

Step 1: In file explorer, your root directory at stable-diffusion-webui. This is the directory containing webui-user.bat.

Step 2: Move the wd-v1-2-full-emma.ckpt or trinart2_step60000.ckpt you downloaded into this folder, without renaming.

Step 4: Copy webui-user.bat and rename the copy to webui-user-waifu or webui-user-trinart respectively.

Step 5: Right click on the copy to edit within notepad and replace its contents with the following, saving your changes.

For webui-user-waifu.bat

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--ckpt wd-v1-2-full-emma.ckpt --medvram --opt-split-attention

git pull
start http://127.0.0.1:7860/

call webui.bat

• Pruning wd-v1-2-full-emma.ckpt to reduce VRAM usage and loading time

  • The Waifu diffusion model is normally 7GB due to redundant training data, but it can be reduced to 3.6GB without any loss of quality
  • Download prune.py
  • Right click on prune.py, click edit, and modify as follows:
    • Change the last line of prune.py to the location of wd-v1-2-full-emma.ckpt on your PC.
  • Open command prompt in this directory, following these steps if you're unsure of how to do so:
    • Make sure file explorer is in the location of prune.py on your system
    • Click on the navigation bar (which should say something along the lines of This PC > OS Disk (C:) > YOUR LOCATION).
    • Clear anything on this bar, then type cmd and hit enter. Command prompt should open in the proper directory folder.

  • Enter the following command:

    python prune.py

  • You will now have a pruned .ckpt.

For webui-user-trinart.bat

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--ckpt trinart2_step60000.ckpt --medvram --opt-split-attention
git pull
start http://127.0.0.1:7860/

call webui.bat