Civitaiトレーニング製 学習元画像は草原での寝ころび床ドン（見た目はほぼ壁ドン） 感覚的に見た感じ、生成結果だけ見るとほぼ同じ見た目では？という感じになってしまった。元々あるタグへの補正誘導型Lora。 ■推奨強度 元々タグとして存在する壁ドン（kabedon)を補正するもののため、低強度でも問題ない 強度傾向 0.3- 壁ドン的な見た目 0.7- 元画像に近い見た目。頭横への手の押し付け感が0.3側より増す やや横長を推奨 ■学習元画像要素 草原での斜め寝ころび床ドン/壁ドン（片手） やや横長 ■その他 ぶっちゃけ寝てても（lyingしてても）、後ろ側が壁なら壁ドンに見えはする。 用途切り替えを意識するなら本来、立ち壁ドンはlying 外すべきかな。とは思う。 Trained on CivitaiA LoRA based on training images depicting a "yukadon" (floor-pinning) pose while lying in a grass field — visually resembling a wall-pinning ("kabedon") pose.At a glance, the generated results look almost identical to the original tags, leading to a perception that there’s little visual difference. This is a guidance-type LoRA intended to enhance the output of already existing tags. ■ Recommended StrengthThis LoRA is intended to reinforce the existing "kabedon" (壁ドン) tag, so even low strength works well.0.3–: Looks like a standard wall-pinning (kabedon) pose0.7–: Closer to the training image, where the hand pressing near the head is more emphasizedA slightly horizontal resolution is recommended. ■ Training Image DetailsDiagonal lying-down yukadon (floor-pinning) / kabedon pose in a grassy field (one hand)Slightly horizontal composition ■ Additional NotesTo be honest, even when lying down, if there’s a wall behind, it can easily be interpreted as a wall-pinning pose.If you want to clearly separate the intended usage, standing wall-pinning scenes should probably avoid using the "lying" tag.

Files with SHA256: bd24d3bf7bb6e3c9ff2f13f72e28501c3c32a42dda1249e2417f2559b20caba9