Lora trained on 40 images of both men & women wearing skull face masks.

Best results seem to be on a strength of (0.7) - (0.08) - Trigger with a phrase such as "man wearing a skull mask". You can also add words such as "balaclava" or specify a type of hat or hood to try and create a full mask with eye holes, although results may vary depending on the model.

This Lora is able to create masks for both genders, and works well with prompts about hair length, as well as NSFW details. Occasionally a mask may appear with a random zipper, so that to your negative prompt to remove it from an image. Unfortunately skulls may also appear at random in an image, and that's my fault for not choosing a unique identifier, but it sometimes looks quite cool.

Note: This is the first Lora I've trained using the standard model rather than strictly realistic. Hopefully this allows for better use in anime and cartoon styles, although testing will be useful.