正面に３つの男性器があり、そこから射精してる瞬間前後をイメージしたコンセプトLoRA

下記辺りを通常のタグ生成より強化

■特徴
・正面３男性器の位置固定（現実で考えるとこんな狭いの無理だが）
・男性器ややぼやけ表現
・精液の量はそれなりに増やしてみた
・射精前、射精中（既に顔に精液付いた状態にしてしまっていたが）、射精後の３シーン切り替え

■シーン切り替え重要トリガータグ
・射精前＝bfecum, open eyes
・射精中＝excum, half-closed eyes
・射精後＝afcum, bukkake, closed eyes
目の開き具合に動作が連動している

■推奨適用強度
0.8- 動作を学習画像に近づける場合
　　低強度にしタグで補正するのも可とは思われる
　　下限 0.4辺り

■その他
・少し動作が混じる時がある。特に射精前。
・arm supportタグは本来射精前の座りポーズ用。これ入れれば混ざりはもう少し抑止されるかもしれないが
　個人的には 射精中楽手画像側に入れてる arms behind back で全部いいのではと思っている

A concept LoRA designed to depict the moments immediately before, during, and after ejaculation from three phalluses positioned at the front.

The following aspects have been enhanced compared to standard tag generation:

■ Features

・Fixed positioning of three frontal phalluses (though such tight spacing would be physically impossible in reality).

・Phalluses rendered with a slight blur effect.

・Increased volume of semen.

・Scene switching between three stages: pre-ejaculation, during ejaculation (note: the model defaults to having semen already on the face during this phase), and post-ejaculation.

■ Key Trigger Tags for Scene Switching

・Pre-ejaculation = bfecum, open eyes

・During ejaculation = excum, half-closed eyes

・Post-ejaculation = afcum, bukkake, closed eyes

(The action is linked to the degree of eye openness.)

■ Recommended Strength

・0.8: To align the action closely with the training images.

(Using a lower strength and adjusting via tags is also a viable option.)

(Lower limit: around 0.4)

■ Other Notes

・Actions may occasionally blend together, particularly during the pre-ejaculation phase.
・The "arm support" tag is technically intended for the sitting pose prior to ejaculation. Including it might help suppress clipping a bit, though...

Personally, I think using "arms behind back"—which I already include for the image where the hand is resting comfortably during ejaculation—would cover everything just fine.