Kasumi from the Dead or Alive fighting game series
Trained from 3D renderings of her in game model
Previews generated with
JANIMA
Better Anime Style Plus
Animated preview using
Smooth Workflow Wan 2.2
Prompt
@animadoakasumi, lapis lazuli outfit, big breasts, cleavage, pelvic curtain, thigh high white stockings, blue leg armor, white neck collar, ponytail, brown eyes,
Notes
Strength: 0.8-1.0
I find using 0.8 helps on some checkpoints to avoid the "3D" appearance because the LoRA is trained on 3D renderings of her character
Tips
add "breasts out" to the prompt to open her clothes
the "blue leg armor" prompt can be removed if you just want her bare stockings.
she has her "wakizashi" trained, but I've actually been using it as a NEGATIVE prompt to avoid it rendering since it sometimes adds the red tassels into generations in random spots.
Enjoy! This is my first attempt ever at training a LoRA, and it's using images I posed and rendered myself in Blender as a dataset. Any tips or pointers are welcome! I just really wanted an Anima LoRA for Kasumi because I really enjoy using Anima when compared to Illustrious, the natural language prompting and consistency is amazing.