This LoRA model is derived from 74 generated images using SDXl (RealVisXL V5.0), I have done testing with a few models, only photo-realistic so far.

Note this is “V3” because I was not happy with the first two versions and did not publish them.

Usage – I find this works best with a weight of 0.6. 

Weights of 0.5 -0.8 are okay, higher than 0.8 seems to distort the results

It seems to produce fairly consistent faces.

Using "black hair" seems to make the hair very black. I have had good results including "dark hair" in the prompts.

All the training images used had brunette/black hair but the LORA appears to be happy with other shades of hair as well. I have tested blue, green, rainbow and blonde.

The model aims to generate a consistent face with a tendency to high cheek bones. It usually generates dark hair

Any resemblance to real persons, living or dead, is purely coincidental and unintentional.