In standard image generation models, the “all fours” prompt often results in images where the arms are fully extended. Additionally, while adding the “top-down bottom-up” prompt can lower the posture, it sometimes results in the figure crouching so low that the head touches the ground.

This LoRA attempts to generate a crawling posture with the elbows on the ground, which is a middle ground between these two extremes.

This LoRA is still in the experimental stage, and currently, it seems to work well only with landscape-oriented images.

肘をついた状態の四つん這いの構図を生成しやすくるうLoRA。

※横長画像でしか上手くいかないかもしれません。