Priscilla from King's Raid.

First off I must say I don't even know how I pulled this off considering I could only find 8 pictures of her with very questionable quality to work on but the result looks decent enough, had to improvise a lot to increase her dataset size, secondly I know nothing about this game actually, decided to give it a try as a challenge when I saw a user asking tips for training characters with too few pictures and then mentioned this character.

Tested with a weight of 0.8, also since her dataset was pretty lacking consider this LoRA a prototype for the time being and some of her original dress features may be missing or inconsistent.

Her main trigger word is kr priscilla, some additional some relevant tags for her are: maid, maid apron, apron, hairband, medium hair, flowers, long braids.

Some other untested tags but used in her dataset training: bow, ribbon, short sleeves, detached sleeves, hair ribbon, frills, maid headdress, hair ornament, bare shoulders.

Check pictures for prompts in general.

Trained mostly on a mixture of both 768x768 and 512x512 images due to the massive lack of images she has but will still perform well at high resolutions with Hires fix, tested with 4x-UltraSharp and 4x-AnimeSharp upscalers, 10 hires steps, 0.2 or 0.4 denoising.

Trained with NAI, images rendered with FoxyaNSFW checkpoint default NAI Vae.