The difference isn't that noticeable so far because the Kotonohas and Yuzuru already had all emotions, but for clients of AI Inc, they only have to record one voice bank now instead of 1 for every additional emotion because the new deep learning calculates that. Also, you can export as an mp3 instead of only a wav file. There is also fusion which lets you combine characters (have the option to use their pitch as a reference or speaking style (ex: Kansai accent) as a reference), for the Kotonohas and Yuzuru, you can give their bud and dark voice all of the extra emotions through voice fusion, too. (I have to test out the software more today, though.)