An AI image of Wakana Umezawa I created from a LoRA trained on 100 copies of the same image.

Since the dataset image has usual AI mistakes, it kinda got reflected in the LoRA model as well, especially the curtain. I used the same base model I used to make the original images.