Text to speech voices tend to be in the dominant dialect and accent of their country — or of their whole language. British English voices are in a posh accent. American English voices are in that weird accent I don’t recognize the location of, but I do recognize as that strange group of people who (falsely) think of themselves as accentless. And so on.
Being half or more Southern in origin, and someone who got speech therapy partly as the result of having what sounded like a nonstandard accent, and got openly mocked by teachers over things like this, this has always pissed me off. Where are all the many and varied accents and dialects throughout the English-speaking world? Where are the accents and dialects considered poor or working-class? In a world where the technology exists to take just about any voice and develop a synthesized version, there is no excuse for not developing voices for a wider range of accents and dialects.
Yes, yes, I’ve heard a million times that people all ought to relish the chance to speak in an accent that “everyone” (read: everyone they think important) understands. That if we want to speak in ways that our families or communities speak, there must be something wrong with us. That we would otherwise appear stupid or uneducated (thanks, and screw you too). I’ve heard all of this.
But I just don’t and can’t buy it. To me all that says is “Please disappear.”
Neither does Dan Bagley, CEO of Cepstral, which produces text to speech voices for Mac, Windows, and Windows CE. He’s from Arkansas, and his company created the voices Dallas (male) and Belle (female), both of which have Southern accents. As far as I know they’re the only widely available Southern text to speech voices.
I was really excited about these voices, but there’s two problems I have with them. One is that they’re available only for Mac at the moment. I hope this will change. The other is a class thing — in keeping with the tendency towards dominant accents even in a non dominant region, they’re definitely not the accents I grew up hearing. And I’m way not into the Southern belle thing.
Fortunately, though, they’re closer to familiar accents than anything else I’ve ever heard. Also fortunately, my current main communication device is a Mac. So I downloaded Belle and have been using it ever since. I have to admit that as far from my family’s accents as she might be, she’s doing better with my natural word patterns than most voices I have used. She’s lower quality than the voices I normally use with Proloquo, but not so much that it would stop me using her. Far better than the old Dectalk voices I used to use.
Cepstral lets people download demo versions of these voices, so if anyone with a Mac is interested I’d definitely try them out before buying. They suggest that too, considering they’re $40 and nonrefundable. The demo versions will randomly say things like “please register me” every few sentences or so, so they’re good for trying out but not for real life use.