This LoRa is an evolution of the "Captive Pose" lora: https://civarchive.com/models/2118478/captive-pose-sdxlillustriousnoobai
Trigger words: 1girl, (naked), (shameful look), (blush), (tears), full lips, (bit gag), (drool), (oiled shiny skin), (arms behind back), (arms restrained behind back), detailed.
Front view.
Three reference poses: standing, squatting & spread legs, kneeling, spread legs and leaning forward with some images.
This lora is not easy to use, depending on the context, some incongruous behavior will occur, and sometimes impossible to remove even when playing with the negative prompts.
I wonder if inserting (arms in the back) between (arms behind back) and (arms restrained behind back) isn't useful, in any case there seem to be better results!
Recommendation for squatting pose: (arms behind back:1.9), (arms in the back:1.9), (arms restrained behind back)
Finally the kneeling/spread legs/leaning forward is the easiest pose.
Doesn't work well with realistic models.
Strength: 0.8 -> 1.2