This model was created on a friend, mostly to see if I could actually do it.

This is made from an actual person, so please keep that in mind when you generate images with it.

From my testing on a few checkpoints, it appears that this LoRA works most accurately at 0.9-0.95. Yes, very narrow parameters. Hopefully I will get better at this as I do it more.

Sometimes the keyword helps, and sometimes it is a hindrance. It mostly seems to depend on the checkpoint used. Experiment with it and have some fun!