Around year 2000, I had tested AT&T Natural Voices and was quite surprised out of the quality of the results that it can yield.
But Acapela seems to be giving AT&T quite a hard time (provided that it’s not in fact the same project). It becomes now very hard to figure out whether some speech is generated or actually recorded from a real voice.
For you to compare:
- Sample A: Sample I generated in UK English with Acapela.
- Sample B: Sample I generated in UK English with AT&T Natural Voices.
However, AT&T developed an INDIAN ENGLISH voice, I cracked up when I generated this couple of samples:
- Sample C: (Serious) Sample INDIAN ENGLISH
- Sample D: And for those whom know Russel Peters… 🙂
- Sample E will be here if you find a Singlish generator! That’d be fun.
Try your own text here: