This is a Lora model of Tamaki Iroha from Magia Record.
There are three variations: transformed, Kamihama uniform, and Takarazaki uniform.
The hair ornament is only visible from the back, so it's optional.
The training image is 1672x957, so please generate it at a larger size, such as 1536x1536 or larger. Generating at smaller sizes like 1024x1024 will degrade the image quality.
The crossbow and Doppelganger activation could not be reproduced. For some reason, the crossbow became a normal crossbow, and although I tried various other captions as suggested by GPT, they either resulted in different weapons or weren't generated at all, so reproduction was not possible. Even trying with just the danbooru tag without Lora didn't work. Doppelganger activation was impossible due to the small number of training images.
Caption used when activating Doppel:
doppel, white face, empty eyes, black void eyes, no pupils, black mouth
After transformation, the face often looks unsatisfactory in full-body shots, so it's recommended to combine it with upper body focus or face focus. The chest plate design changes when the hood is removed, but this is due to the lack of training images without the hood. The low ponytail may sometimes appear outside the hood, but this is how the danbooru tag itself is trained, so you just have to accept it.
When making the hair long, you need to include hair down, otherwise the hairstyle won't change easily. Braids will appear quite often, so add braid to the negative prompt. Even then, braids will still appear at a low probability.
Since there are few full-body images of the Takarazaki uniform, lowering the intensity will make the lower part of the uniform and skirt look like the Kamihama uniform, so those who are particular about accuracy should keep the intensity at 1.
The trigger word and caption have a spelling mistake where "shoes" is written as "shose," but there is no difference in the generated output regardless of which one is used.