The LoRA isn't trained on what's inside the gag but if you know, you know.

Triggers are " gagged" and "leather panel gag."

This one's a bit less flexible and more finnicky than some others. Strength around 0.8 works best.

If the image comes out blurry, try lowering LoRA strength and adding parentheses around the LoRA trigger, lowering CFG, or both.

Due to the dataset, this model really likes to make leather collars even if you don't ask it to. It may also generate pasties/tape. Could be a plus or a minus depending on taste.