Civitai POV 見上げ式被さり体位＋半透明　/ POV looking up face-to-face position with girl on top. plus invisible man.　NSFW

カテゴリ修正

正しい言い方が分からない。

足を広げてのまたぎなしの少女載せ、被さり体位。半透明男性を添えて～

ある意味、invisibleシリーズ。

既作成済のinvisible leg lock に似た要素はある。

■本来の目的

ベッドおよび男性を透明にし、少女を男性の上に載せ、少女と天井を見る構図としたかった。

透明を複数重ねるといい事は全く無いので透明ベッドは途中でやめた。

現状だと生成の際、余計なものを反映してしまうのがどうしても回避出来ない（control net使えばいい気はするが）ため、少女と天井を見る構図は断念

青空ではそれらしくなったが、あまりヒット率が良くない。

■invisible leg lock v2.0に比べ

・上で書いてる通り、目的としてる位置方向と背景の組み合わせのズレが出やすい
　うまく表現しきれてないと言えばそれまでだが男女が寝てる状態で性行為してる。
　というよりは男女とも立ちながら性行為してるように見え、背景もそれに合わせられてしまう事が発生しやすいのではないかと思われる
　→つまり、背景描写が狙い通りにならない場面が多くなる想定

・半透明男性の色の薄さの安定感は少し向上。別キャラとか入れ始めると崩れやすくなるが

・半透明男性の色の薄さの安定感に伴い、Adetelarの全体掛けがある程度安定する

・半透明男性の頭の横位置があまり安定しない

・半透明男性の奥行き配置は相変わらずガチャ。leg lockに比べれば少し安定した感じはするが

・縦長推奨

・必要プロンプトが多い。実質的にLoraが補助的機能に留まってしまっている

・Lora強度の推奨範囲が非常に狭い感じがある

外部Lora含め、キャラ変可能は確認済。

推奨強度 1.0前後

Straddling-Free Girl-on-Top Overlay Pose, with Semi-Transparent Male – A Variant of the "Invisible" Series

In some sense, this is part of the invisible series.

It shares some conceptual elements with the previously released invisible leg lock.

■ Original Concept

The initial goal was to place a girl on top of a fully transparent male and bed, viewed from a ceiling-facing perspective as if the girl and viewer were looking upward together.

However, stacking multiple transparent elements (bed + male) proved highly unstable, so the transparent bed idea was dropped midway.

Even now, when generating this setup, unintended elements tend to appear in the composition—perhaps ControlNet could solve this—but I ultimately abandoned the ceiling-view concept.

Using a blue sky as the background created a decent illusion, but the hit rate remained low.

■ Compared to Invisible Leg Lock v2.0

As noted above, it's prone to composition drift: the character placement and background often fail to align as intended.

Instead of appearing as a couple lying down in bed, the result frequently looks like they're standing during intercourse, and the background adapts accordingly.

→ So expect more inconsistency in background interpretation.

The color stability of the semi-transparent male has slightly improved. However, introducing additional characters can easily break this.

Due to improved transparency stability, applying ADetailer across the entire image works more reliably than before.

Horizontal positioning of the semi-transparent male's head remains inconsistent.

Depth layering of the male is still a gamble. Compared to leg lock, there's marginal improvement.

Vertical layout is recommended.

Requires many prompt elements; as such, the LoRA ends up serving more as a supplementary tool than a standalone solution.

The optimal LoRA strength range is extremely narrow.

Character switching, including use with external LoRAs, has been confirmed to work.

Recommended Strength: Around 1.0