This is my first LORA trying to do something I'd liked.

Trained using a combination of generated and non-generated images

Weight around 0.8 should be fine.