this lora is more flexible with prompt
model of nina kosaka from nijisanji en, trained with about 100 images
some of the image is made by Rakosz
