This lora is made from pathetic pose of Asuka, flipped pathetic pose from Simpsons, and a few "forward style" pose where character in the middle. There was a few realist picture in dataset, but most of them is "fan parody", so it preferable to use 2d/anime styled base models.The Lora itself is "fryed" so you probably will have to find your own perfect value for each character. For example(of being fryed) it coudn't generate normal symbol shaped pupils for hu tao. So maybe you should use it two samplers (for comfyUI) with this lora (doing main job) and without but with control net from the from the first generation (adding the details). Version with less amount of epoch is worse generate this concept.Work best with girls models (lora models) (but still not with everybody). Work worse with male most of the time. Once accidentely made work with Shinji lora.It is very "dirty" in terms of the style variety in dataset, so it will generate with it's own "style"(maybe will fix it in future through generating images with another style but with the same concept)...In general this whole lora looks like tag "from below", but the character has also some titled to the side.Trigger words: "pathetic \(meme\, patheticasukastyle, patheticsimpsonstyle, patheticfrontlstyle" (that for each type. I usually put patheticfrontlstyle in negative to make it "only titled") and possibly this one - PatheticTestEight. For disgusted expression you can use next set - "(very angry, disappointed:1.1),"

Files with SHA256: b0d9d5a377054496bda446e9b26438becf7036d057808162903e30edc5927e0e