Illustrious first attempt. <ore tags in training than pervious character model
more tags same dataset
