Looks good, will have to try this out in the next few days. One question though, I noticed that the results look a bit fake in some images, almost like it's cheap makeup effects. Are you training all on original images or are you using some AI images? When I was trying to improve my body horror lora I decided to include a bunch of images I generated myself as finding good reference images for training this kind of style is difficult. My results were coming out very simular to yours where a lot of generations just lacked realism, even when using realistic models. After several tries I ended up just scraping the idea of using AI images completely.