<a target="_blank" rel="ugc" href="https://civarchive.com/models/2453268/civitai-shadow-play-shoji-and-girl">Civitai 影絵-障子と女性 / Shadow Play - Shoji and girl. - v1.0 | Illustrious LoRA | Civitai</a> のNSFW性行為LoRAテスト使用例 多くのコンセプトLoRA類の場合、うまく出力できない際、目的コンセプトがだんだん崩れていくようなケースが多いかと思っているが 当LoRAの場合、Seed値によってコンセプト部分が全く機能しない出力がいきなり出たりするのが難点。 特に元モデルページでも書いた通り、白黒二色に完全に抑えるという事が出来ない。 NSFWテスト。 モデル側学習元の影響が強いのか性行為時は特に色を塗られやすく 上で書いた通り色が抑えられない事が割と出る all foursは元々組み込んでたので比較的安定してると思うが他の追従性、発生率はそれなりといった所か 添付ワイルドカードはテスト用性行為体位<a target="_blank" rel="ugc" href="https://civarchive.com/models/2453268/civitai-shadow-play-shoji-and-girl">Civitai 影絵-障子と女性 / Shadow Play - Shoji and girl. - v1.0 | Illustrious LoRA | Civitai</a> NSFW Sexual Activity LoRA Test Usage ExamplesWith many concept LoRAs, I think that when output fails, the intended concept gradually collapses.In this LoRA, the difficulty is that depending on the seed value, the output may suddenly become completely non-functional.As noted in the original model page, it is particularly difficult to completely limit the color to black and white.NSFW test.Perhaps due to the strong influence of the model's learning source, the color is particularly easily overlaid during sexual acts.As noted above, the color often fails to be suppressed.All fours was built in from the beginning, so I think it's relatively stable, but the other tracking and occurrence rates are only so-so. The attached wildcard is a test sexual position

Files with SHA256: e96c89cd27bce27cca55780bf6779c67bbe7558f20d6ae7badf70526dacf2491