Kasumi from the Dead or Alive fighting game seriesTrained from 3D renderings of her in game modelPreviews generated withJANIMA<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/2642932/janima">https://civarchive.com/models/2642932/janima</a></li></ul>Better Anime Style Plus<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/2675648/bas-better-anime-style-plus-or-anima-base-v10-or-model-or-checkpoint?modelVersionId=3004325">https://civarchive.com/models/2675648/bas-better-anime-style-plus-or-anima-base-v10-or-model-or-checkpoint?modelVersionId=3004325</a></li></ul>Animated preview usingSmooth Workflow Wan 2.2<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/1847730/smooth-workflow-wan-22-i2vt2vfirst2last-framemmaudio">https://civarchive.com/models/1847730/smooth-workflow-wan-22-i2vt2vfirst2last-framemmaudio</a></li></ul>Prompt@animadoakasumi, lapis lazuli outfit, big breasts, cleavage, pelvic curtain, thigh high white stockings, blue leg armor, white neck collar, ponytail, brown eyes,NotesStrength: 0.8-1.0<ul><li>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</li></ul>Tips<ul><li>add "breasts out" to the prompt to open her clothes</li><li>the "blue leg armor" prompt can be removed if you just want her bare stockings.</li><li>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.</li></ul>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.

Files with SHA256: 7140848ed4bb0f5db875fa5af1d5df3477887b64f002a2e307386e27e310d5c0