A few days ago I came across an artist that drew really interesting monster girls and immediately wanted to train the style.

The dataset is, ofc, almost entirely made up of monster girls, so naturally that's what it'll provide the best of. Most of their art is also full body concept art against a simple background, so I recommend using the full body tag among other body descriptions to achieve that, plus adetailer or normal hires fix if faces turn out bad.

It might not work perfectly right out the gate, but mess with it enough and you can get some really good results out of it. The models that gave the best results were primarily anime styles, namely meinamixV10 and darkalfaEffortless_v11. I also recommend referring to the preview prompts if you need help with describing certain things or getting a certain look.

ISSUES

• Images tend to generate with particles or random spots all over the place, or the background will have some odd-looking texture to it. However, this has been a problem I just recently started to encounter with other LORAs I've done as well, so it could be just me. I plan to update this if I can manage to fix that.

NSFW capable, but nothing hardcore. Trigger tokens aren't strictly necessary, but I still add monster girl to the beginning of the prompt somewhere since it was a main token during training.

Images you make with this model and what you do with them are your responsibility!