Looks cool. I feel like it might be best to default to google's text-to-speech API if a connection is detected - in the example video, the windows t2s pronuncia

Looks cool. I feel like it might be best to default to google's text-to-speech API if a connection is detected - in the example video, the windows t2s pronunciation sounded like "plum" and the user wouldn't have any other way to know it was pronounced incorrectly until they cross-check with google or spell the word incorrect. After a few mis-pronunciations they'll probably just use the google version every time anyways. Not that it's that big of a deal or anything.