This LoRa aims to emulate Ie's artstyle.The trigger is ie2koma artstyle.This artist mostly specializes in "instant loss" scenarios, usually in a 2koma, comic format. It kinda works in the model, too. Like 50% or so?It does tend to produce stray speech bubbles, particularly empty speech bubbles along with it, so negative prompt those. Also, the base material was censored, so put that in negative, too.

Ie artstyle (Pixiv, etc) [Pony]

some tag optimization was necessary in the dataset

Files with SHA256: b529332dfb1f5e61be43b3b4ad2a01677af5bef344ff10b22d11340bc404c555