It was trained exclusively on images of this artist.
The Lora is around 65% accurate, but it works okay.
A better version could be made later.