Changes in v.2 can be seen on the right by clicking "About this version"
I tried to make model that excels in cuteness with cold vibrant darker colors and ability to do nice darker images. It has "enhanced" hair and the hair can really wild sometimes
(～￣▽￣)～
The mix likes blue and has tendency to lean on it.
Preview images were done in comfy and you might get different results in A1111 or any other UI. Comfy is not fond of giving accurate metadata so it is missing some information.
Has basic MSE 840000 vae baked in. VAE is pretty much preference so i recommend using which ever you like.
Sampling settings come to preference and have not noticed any bad samplers so far. Most likely excels best with:
Sampler: DPM++ 2M Karras, euler ancestral, SDE 2,
Clip skip: 2
cfg scale: 7
Basic anime prompts work:
masterpiece, best quality,
worst quality, low quality,
etc...
This is a merge between:
and several loras i have trained.
Thanks for @Anzhc coming up with the fantastic name.
You are solely responsible for any legal liability resulting from unethical use of this model!
Follow civitai tos when posting images!
Description
Tried to make characters look less childish while still trying to maintain the cuteness.
Toned down the wild hair a bit, it was being too wild sometimes.
Follows the prompt more without adding too much random detail. Sometimes the V1 added unnecessary detail.
Backgrounds should be better.
Colors should follow a bit more precisely now, toned down the "blue" tint a bit.
Merge models are the same