Doemutan Style Illustrious

Doemutan style, trigger word is doemutan

My first Lora, trained on 100 images with base illustrious XL model, a lot of which are greyscale managa images which have some speech bubble in, mosaic censoring and text, I have tried to clean up the data set as much as I can but its not perfect. If outputs have a lot of text or censoring try putting text, censor / mosaic censor and speech bubble in negative prompt. Any feedback would be appreciated

grey version seems to be better at making greyscale/manga images, but has worse colors when generating images in color, V1 seems to be better at generating color images.