In standard image generation models, “side-by-side” prompts generally work well. However, the composition may break when specific compositions or prompts are included.

This LoRA assists in generating scenes where men and women are seated on sofas, benches, beds, etc.

Placing the “couple” tag in the positive word list makes the woman more proactive, while placing it in the negative word list makes her more passive.

男女が隣り合って座る構図を出しやすくするLoRA

"couple"タグのポジティブワードに入れると女性が積極的になり、ネガティブワードに入れると女性が受動的になるようにデータを工夫しています。