I only used anime screenshots for the dataset, but for whatever reason I couldn't really capture the look of the anime like I wanted to.
