This LoRA reproduces decent hand over mouth scenes.
The model currently has some limitations:
There is only one pose (from behind).
For NSFW content, it needs more tags or slightly increased NSFW tag weights.
But now the good things:
V2 has a success rate of approx. 80% (tested with AutismMix confetti)
No overfit, any creature can cover the mouth of any creature.
Supported descriptions for the person covering the mouth and the victim.
Prompting tips:
The main trigger is "handgag", and the two persons in the scene: "female to female", "male to female", "female to male", "male to male".
To describe the person who is covering the mouth add "handgagger" after the tag. Examples: "ginger handgagger, blue eyes handgagger, red jacket handgagger" etc.
The same goes for the victim, such as "white shirt victim, brown hair victim, surprised victim" etc.
You can also describe the hand with "small hand, big hand, strong hand, red nails, blue nails" etc.
The model also supports gloves, the main trigger is "gloved". But you can also describe the glove like "leather glove, black glove, fingerless glove", etc.
The dataset includes various environments and humanoids, so it won't struggle to create robots, furry, aliens, ...
Use tips:
20 to 30 steps max. Too many steps will result in increased risks of extra fingers.
FaceDetailer can improve the thin details of the hand like nails, rings, etc.
Using a tiled upscaler will produce better results.
Please use this LoRA responsibly following ethical and moral values. I take no responsibility for its misuse.
