Cool! I'm not great at Chinese but I have to speak slowly for it to recognize the tones/words. I wonder how fast the training data is.