This is my second lora release. I still have no idea what I'm doing; the results will vary. Please be nice.
All preview images were cherry-picked but unedited; no touch-ups were done in Photoshop, nor were hires fix, img2img, or inpaint used. Obviously, you'll get better results with those tools.
This is the Beauty trainer class from Pokémon X/Y. The LoRA seems quite stable at full strength, but I recommend adding 'solo' to your prompt, as it otherwise tends to clone her. For her odd hip ornament, I used the term 'medallion belt' in the training data, which seems to work well enough. The commonly used terms in the training data include 'XY beauty, brown hair, updo, single bun hair, sidelocks, eyewear on head, sunglasses, spaghetti strap, denim, jeans, medallion belt, high heels.'
Her clothing and body type can be easily modified through prompting. It should also work fine for NSFW purposes, although for some reason, when pubic hair is prompted, it tends to produce a lot of very thick pubic hair. In my limited testing it sometimes gets a little unstable doing NSFW. I will have to test more later. Lowering the strength of the lora should help alleviate this.
My usual settings for testing this have been Euler A at around 30 steps with a CFG scale of 7.