First attempt at creating an NSFW action LoRA for Moss Image. This is a difficult model to train, so consider this LoRA an experiment.

The main issue is that it tends to generate the same face (particularly the nose and certain facial features) and relies heavily on having a good prompt.

The good thing is that it's 100% compatible with character LoRAs, and in fact, it works even better when combined with them than when used with the base model alone.

Workflow in the images (standard Moss base model + 8-steps DMD2 LoRA + this Lora).