Stable Diffusion 1.5 mixed with Waifu Diffusion, and the NovelAI leak has me making big tiddy goth GFs on command. Lotta nice ladies in fancy dresses, too, but that's besides the point.
Use this webui. It'll auto-update every time you run webui.sh.
Move the checkpoints you wish to use into the 'models' folder
Start webui.cmd (windows) or webui.sh (everyone else)
Go to 'Checkpoint Merger' tab
Choose the checkpoints. takes shitloads of RAM to load multiples
Choose the percentage of Model B to mix with Model A
Do this a few times with whatever models you like
OPTIONAL: embeddings exist to even further constrain and improve
PRO-TIP: do NOT use negative prompts when getting your first-pass images. It fucks up the poses. Depends on model how severe, my mix seems largely unaffected
This is like AI art.
The new algorithm is nuts.
Stable Diffusion 1.5 mixed with Waifu Diffusion, and the NovelAI leak has me making big tiddy goth GFs on command. Lotta nice ladies in fancy dresses, too, but that's besides the point.
How do you mix em?
https://github.com/AUTOMATIC1111/stable-diffusion-webui