V3:
Prompts: cinnamonroll, blue eyes, white body, furry, chibi, feral, blush,
optional: anime screencap, traditional media \(artwork\)
Lora strength: 0.9~1.1
Negative: eyebrows, pawprints, multiple ears
My original plan with V3 was replacing all the dataset with screenshots from the movie, I hate to say this but I got lazy and just added a few images to the old dataset. Training this lora was a bit of a pain (some IRL stuff) but I think it was worth it. The Illustrious version is going to be released at a later date.
If the lora is consistently generating a nose you should remove the "furry" tag, that seems to do the trick. (none of the images used have noses).
I used a few screenshots from the anime (like 3) and a few hand drawn one (sorry about that) and using one of the tags in optional can alter the style a bit.
V2: Looking back I should have worked a bit harder on cinnamon roll, so after a protect I was working on kind of fell apart (the characters I wanted to make already had a lora), I decided to fix this one instead, I hope you guys enjoy it. I tested V1 in Hoseki but V2 was tested in Hassaku, you are free to use any model you want but just keep that in mind.
V1: I got a request for Cinnamoroll, I was going to leave him for later but why wait? During testing I used the Hoseki - LustrousMix checkpoint, Going to be honest with you guys, Cinnamoroll is among my favorites but I like Wish me mell or My melody way more.
Hope you guys enjoy it and remember, any feedback is appreciated.