My All in one character LoRA for Alicesoft's Oyako Rankan - now ported to Illustrious. I have also included Kyouko into the training dataset this time and some support for prompting multiple characters at once. Very unreliable results for multi character prompting however, so inpainting is still recommended to clean it up. It does serve as a good base though since you do not need to make major changes with inpainting due to getting close with the initial prompt usually.
You must enter min-naraken into the prompt if you want to replicate the artist style and character outfits accurately.
Example without inpainting:
Prompt: <lora:Oyako_Rankan_Illustrious:0.8>,3girls,saya and kotone,large breasts,ririko,medium breasts,min-naraken,indoors,full body,standing,seductive smile,group picture,casual,
Below is the same image with improvements made by inpainting.
Three passes over the image, each inpainting one character using the following settings.
prompt template: <lora:Oyako_Rankan_Illustrious:0.8>,1girl, <character name>, <breast size>, min-naraken
Soft Inpainting enabled
ADetailer for Hands and Eyes enabled
Only Inpaint Masked
Denoise of 0.6
Upscaled with R-ERSGN 4x+ Anime6B
Description
Use min-naraken in prompt for style. See trigger words for how to prompt characters and outfits. Initial release and new dataset from my Pony version of this LoRA - there has been zero fine tuning of the dataset so far.