The vocal “swing” is just the slightest bit off, but the vocal quality is so good you can largely ignore it. If you stick with it long enough, there are actual glitches in the words.
It turns out there are some serious problems with the actual accuracy of the show content, but that’s almost incidental.
One of us should figure out how he/she is doing that.
I found a video by Nilanjan Nag where he uses what seems to be his real voice. I’m guessing he’s Indian (continent, not Native American) and has an odd microphone technique. So I’m not shocked he decided to use Text to Speech. It was probably easier than anything else he was doing.
I agree that it is text to speech, but it is pretty good text to speech. If you find out, I’d be interested to know which text to speech engine he is using.
“Neospeech” is probably not the one being used, but it is pretty good:
I was thinking about that. It might pass, not because of general shoddy QC, but from overwork. Every third person on earth is trying to read audiobooks (make a mint and retire to a seaside cottage on Mal de Mer, etc.)
I was a good way into the performance before I noticed anything odd. If I was having a really bad day and had just written rejection notices to forty readers all of whom submitted cellphone voices, I could welcome a computer voice as being an island of good quality in my sea of garbage.
“I finally got rid of my distortion by applying Noise Reduction three times…”