<h1 id="heading-971">Update v2.0</h1> Completely retrained on a smaller and more curated dataset, which means much better accuracy and consistency, with simpler prompting, at the cost of a bit of flexibility, and a bit more dependence on the checkpoint you use. Overall, I like this version much better, and if you liked my newer Loras, this will be more like those. Weight around 0.7 worked best for me. The activation tag is shizuka masou, and this time it should get her physical appearance without additional descriptive tags neccessary. For her usual clothing: cloak, yellow thigh boots, witch hat was generally enough for me.<h1 id="heading-972">v1.0</h1> My favorite girl from the Rance games. This is my first LoRA, so it's nowhere near perfect, but it can recreate her clothes and general appearance pretty accurately, so I'm happy with it enough to share it. Also works for NSFW. Weight ~0.5 is what I had the most success with. It has the activation tag shizuka masou, but it's best to also add the descriptive tags: green hair, very long hair, blue eyes. Also, for her clothing: yellow boots, thigh boots, belt, coat, witch hat. I really like how this combines with this <a target="_blank" rel="ugc" href="https://civarchive.com/models/4982/anime-screencap-style-lora">Anime Screencap Style LoRA </a>as you can see in some of the sample images.

Files with SHA256: fabb2f1e572efc0db39e1ef0eebc4c3fe1699b8e31097bc57519313dc7c6a61a