SDSetup

Setup your Stable Diffusion web UI repository by looking at files and civitai model links sent in the Discord channel of you choice.

Disclaimer: for now I use this repo as a way to have it uploaded and online and keep track of the progress, commits are by no means "good commits".

Installation and Running

Download the setup.py and config.json files or just run:

git clone https://github.com/lakazatong/SDSetup

Fill up config.json with the required fields and your preferences, then run:

python setup.py -m

Your setup.py is now ready!

Put it in any Stable Diffusion web UI repository and run:

python setup.py

Your Stable Diffusion web UI repository is now setup!

Instructions and Tips

• Messages with only text (no attachments, no embeds) are SKIPPED

• Use reactions to give instructions in the discord channel : 🚫 SKIP, ❌ DELETE (prioritizes SKIP over DELETE if both reaction are on).

• Deleting a file/model or removing a model from your civitai favorites will delete it.

• Some good to have files are in the files folder, you can upload them in your Discord channel if you like.

• Some interesting lines in ui-config.json (for A1111 release 1.5.2):

  txt2img:

    3: default positive prompt
    5: default negative prompt
    37: default sampling method
    39: default sampling steps
    83: default width
    88: default height
    94: default batch count
    99: default batch size
    104: default CFG scale
    109: default seed
  

  img2img:

    189: default positive prompt
    191: default negative prompt
    240: default resize mode
    262: default sampling method
    264: default sampling steps
    273: default width
    278: default height
    295: default batch count
    300: default batch size
    305: default CFG scale
    314: default denoising strength
    319: default seed
  

Warnings

• This script do not check if models were manually installed before the first setup, because models file names could have been renamed thus giving no information on what model it is. It could not even know if it's on civitai (a -c, --cache option could be added to seek for installed models and update cache with what it finds but sounds not reliable at all). So if you want all your previous models to be managed by this script as well, delete them all and let it download them setting up its cache.

• The only right way to reset the cache is to delete the .setup-cache cache file.

• Do not rename downloaded models, it can cause issues.

FAQ

• What's 'discord_auth_token' and how can I obtain mine?

Go to https://discord.com/app in your browser, login if not already, open the Developer Tools by pressing f12, go to the Network tab, make sure it is recording traffic by noticing:

on Chrome / Firefox

Click on any channel from any server, press ctrl+f, enter 'Authorization' in the Search field that opened on the left, you should get something like this:

The 'discord_auth_token' you are looking for is in these files, if you click on one you will get redirected to the file, scroll down to Request Headers, it will be highlighted in yellow c:

• What's 'civitai_api_key' and how can I obtain mine?

Go to https://civitai.com/user/account, login, scroll down, in 'API Keys' click 'Add API key', enter the name you want and copy the key you got, that's it c:

• How can I get the ID of a Discord channel?

Right click on the channel and click 'Copy Channel ID' c:

TODO

• Add a way to rename downloaded models without breaking everything lol (updating the cache basically, could be done with smth like python setup.py -r <path to model> new_name (with or without the extension, using the path to be able to use TAB so that it's easier).

• Add support for all types.

• Have a preferred sampling method for each checkpoint (guess it based on the one used in the previews and prompt the user or use a default one if no preview).

• Cache settings after each txt2img generation in case it crashes and load them after reloading.

• Close the runpod after ~5 mins and terminate it after ~10 mins of inactivity.