The model I think is fairly self explanatory.Probably should have trimmed down the doujin title to just one language, does anyone know if length of tags or names of things in how many characters they are affects how they're tokenized? Works fine at a strength of 1. Previews made with novaFurryXL_illustriousV20.