Experimental lora trained on images of male facials.
Seems to work fine at weights 0.7-0.9.
Works best with img2img and inpainting.