This is a LoRA (technically LyCORIS) for PonyXL aimed at making Bondage / Suspension / Shibari / BDSM more accurate. PonyXL models can generally generate these concepts, but I've found that they do so at the sacrifice of prompt adherence, i.e. prompting "suspension" will all too often make the subject nude, even if the prompt specifies otherwise.
This LoRA is trained with the activation prompt sts_suspension (it was initially a LoRA for suspension, but it's generally good for just about anything BDSM). The concepts that you can prompt with this LoRA are: suspension, bondage, rope, shibari, BDSM, gag/ gagged (although be careful with gag prompts). The LoRA also generally responds well to camera positioning prompts such as, from below, from_side, from above, and straight-on. This LoRA does have training data for upside-down, but I've found SD models to be quite finicky with such subject positioning.
Recommended weights are 0.4 to 0.7. I mostly use 0.5 or 0.6 myself. Example prompt:
score 9, score 8 up, score 7 up, (source anime), 1girl, bdsm, green hair, high ponytail, blue eyes, solo, torn clothes, bottomless, *****, ***** juice, sts_suspension, from below, bondage, bound, red rope, suspension, hanging, (arms up, legs up, legs together), dungeon, highres, absurdres, masterpiece, best quality, ultra-detailed, illustration,colorful, masterpiece, best quality, ray tracing, 8k, extremely detailed, soft lighting, tracing, soft shadows, detailed face, ->LORA-GOES-HERE<-