Lora to reproduce style from the succubus doujin by Fue, had 4 included character.
Description
FAQ
Comments (7)
Amazing, thanks. I wanted one of this work for a long time.
It doesn't really work as well as I'd like to but the doujin is quite hard to use as dataset, it's quite messy :/
I know, monochrome datasets can feel very hit or miss. You train them too little and seem they can replicate the source material very well but are not flexible at all, or you train them too much and are flexible but the style gets muddied. Yeah using doujin is a pain to get a good dataset. Which is why I get lazy and just put things off for so long.
lol, I had the uncleaned dataset for over a year already, only decided to used it last week
@secretpablo Im sitting on so many that I wouldnt know where to start. BTW wanted to ask, do you try to clean backgrounds, or just remove the remains of other panels that are left on the cropped image and such?
@PinGOrio I do both usually, depending on the amount of effort I want to spend on a lora it varies but I always try to make the subject as clean as possible.
nice, good lora