This LoRA heavily learned the show's style. For that reason I recommend using a weight of 0.8 for most prompts. Higher weights and it becomes less flexible stylistically. If you do want to mimic the show's style then go full weight.

The training data for V1 is entirely made up of screenshots of season 1. Mainly because his looks change slightly in the later seasons with scars and scratches etc. I've also only watched season 1 as of now (please don't hate me). If I have time someday I'll experiment with introducing fanart to the dataset to broaden its style range. For now, the looks will be faithful to how he is depicted in the show