Elfina & Ann from H anime "Elfina: Yoru e to Urareta Oukoku de... The Animation (エルフィーナ～?夜（ヨル）へと売られた王国で…～)".

This model offers a mini-sized version of the original art style.

You could use the prompts below to create characters:

Elfina

1girl,mai_ds,mai_ds_clea,short_hair,blue_eyes,blonde_hair,hair_between_eyes,medium_breasts,armor,black cloths,shoulder_armor,pauldrons,breastplate,blue_skirt,armored_boots,holding sword

Ann

1girl,mai_ds_nara,breasts,red_hair,short_hair,hair_between_eyes,animal_ears,cat_ears,animal_ear_fluff,red_eyes,collarbone,navel,long_sleeves,black_jacket,open_clothes,bandeau,black_shorts,belt,midriff,thighhighs,thigh_strap,thigh_boots,black_footwear,holding polearm

Recommend preferred Strength: 0.7~1.

The blonde princess's joyfully popular experience of a fallen kingdom... is also an obscure OVA.

Someone asked whether a set of extension matrices (LoRA) can include multiple characters or even different styles. The answer is yes, but it requires more careful labeling of the dataset, and since the materials may increase, the computing resources needed for training might also be greater. I personally believe that there is no superiority between combined LoRA and single LoRA. I prefer to train character data separately, which may take more training time but makes it easier to organize labeling and handle errors. :)