「penalty buck/spanking stand」が好評でしたので、定番の「Bondage Bench」あるいは「Spanking Bench」や「Restraint Sex Bench」のデータを作成しました。前作は床置き式ですか、今回の拘束台は空中に固定された姿勢になります
トリガーワードは「bondagebench」、推奨プロンプトは「stationary restraints,all fours,bound arms,bound legs,bound body,arm rest,harness」あたりになります。強度を下げるとアームレストとボディーハーネスは再現時に割と抜ける傾向もあります。
視点は比較的どこから見る形に安定した再現はできていると思います。でも真正面は若干苦手かな? と。
サンプル画像はSFW仕様ですが、NSFWの画像も「nude」や「pussy,anus」などのプロンプトでは出ます。異物挿入系や性交関係も安定して組み合わせられると思います(基本の体勢がall foursなので……)
変則的なプロンプトとしては「bondagebench,stationary restraints,kneeling,arms behind back,leaning back,bound arms,bound legs,bound body,harness」や「bondagebench,stationary restraints,on back,arched back,bound legs,bound body,harness」によって、作例にもあるような膝立ち・仰向け拘束台なども出せる場合があります(不安定ですが)。ポーズと拘束台の再現の追随がどうなるかも実験してみてください。
創意工夫のある作例をお待ちしております。
Since the “penalty buck/spanking stand” was well-received, I've created data for the standard “Bondage Bench,” “Spanking Bench,” and “Restraint Sex Bench.” While the previous version was floor-mounted, this restraint bench positions the subject suspended in mid-air.
The trigger word is “bondagebench”. Recommended prompts include “stationary restraints, all fours, bound arms, bound legs, bound body, arm rest, harness”. Lowering the intensity tends to make the arm rests and body harness less stable during recreation.
I believe the viewpoint is relatively stable and reproducible from most angles. However, the direct front view might be a bit tricky?
The sample images are SFW, but NSFW images can be generated using prompts like “nude” or “pussy, anus”. Insertion and intercourse elements should also combine reliably (since the base position is all fours...).
For unconventional prompts, combinations like “bondagebench, stationary restraints, kneeling, arms behind back, leaning back, bound arms, bound legs, bound body, harness” or “bondagebench, stationary restraints, on back, arched back, bound legs, bound body, harness” can sometimes produce kneeling or supine restraint bench poses seen in examples (though unstable). Please experiment to see how the pose and restraint bench replication follow through.
We look forward to seeing your creative examples.
Description
FAQ
Comments (4)
The pose itself works very well, and it gives me what I want as far as that every time. However even using WAI it seems to want to produce more realistic images and its a little tough to stop it. I put "realistic, realism, photorealistic" in negatives. Any advice?
Since the source data is real-world footage, higher strength settings tend to produce more realistic results. Even with WAI, using positive prompts like “anime coloring” or “anime screenshot” should allow for more 2D-oriented generation while maintaining accurac
@aramoto Thanks! I'll give that a shot.
To use LORAs that are based on realistic images, it's indeed possible to 'force' the AI to create anime / 2D images, but you need tricks like lowering weight (e.g. 0.5) and then INCREASING tag weight, e.g. (bondage:bench1.3)






