V1: Trained on 68 QwenAnime "de-glossed" portrait pictures.
Trained with "realistic", aesthetic Anime portrait images from Qwen.