Trained on novelai on 96 images (1024x1024) of Secre Swallowtail with network rank and dim: 32/16. More flexible and more accurate now than older version. It works without using the trigger word, but I recommend using it anyway.
Use red eyes if eye color gets wrong.