Trained with better captions and more images
Trained on 230 images. Doesn't know characters that well, probably needs better captions.
