Home › Blogs › figleaf's blog

Her Master's Voice? Apple's Oddly Gendered Text-To-Speech Samples

Fri, 2008-05-02 08:06 — figleaf

So last night I wrote a “Who Knew?” post about using Apple’s Text to Voice feature for proofreading. Since I’m a relative newcomer to the Mac I followed the technology blogger’s advice and installed “Alex,” the newest, most natural-sounding voice and just started using it without fanfare.

In comments though, Zeborah checked out all the available voices and…

I’ve had my text-to-speech set to Vicki for a long time. Normally it’s little use to me — I read way faster than I listen so it drives me batty to have to wait — but it comes in handy when eg I’ve got the flu and my eyes are hurting.

So, I go to listen to Alex and get side-tracked into listening to the sample sentences spoken by the various characters. For the novelty ones, the sentence generally has something to do with the kind of voice:

Bad News: “The light you see at the end of the tunnel is the headlight of a fast approaching train.”
Bahh: “Do not pull the wool over my eyes.”
Trinoids: “We cannot communicate with these carbon units.”

The male voices also have sentences expressing their personality or something about them:

Albert: “I have a frog in my throat. No, I mean a real frog!”
Alex: “Hi. I’m a new voice for Leopard.”
Bruce: “I sure like being inside this fancy computer.”
Fred: “I sure like being inside this fancy computer.”
Junior: “My favourite food is pizza.”
Ralph: “The sum of the [...] square of the hypotenuse.”

And then there’s the female ones, which, with the exception of Princess, have no personality and no care for anything except their owner:

Agnes: “Isn’t it nice to have a computer that will talk to you?”
Kathy: “Isn’t it nice to have a computer that will talk to you?”
Princess: “When I grow up, I’m going to be a scientist.”
Vicki: “Isn’t it nice to have a computer that will talk to you?”
Victoria: “Isn’t it nice to have a computer that will talk to you?”

Fascinating.

Quoted with permission from here.

Being the literalist I sometimes am around technology probably woudn’t have tried that experiment, so I can’t take credit. And a little quick Googling suggests no one else has noticed either though I’ll obviously update the post if it’s been mentioned elsewhere. Anyway, till then Zeborah ought to get the nod.

In a follow-up comment to the same post she generously added that

It’s probably fair to point out that these voices have been created one or a few at a time over… oh, I can remember some going back ten years at least, I’m sure. Possibly the relevant software engineers have never sat down and played the test files one after another. Even so, someone(s) made some choice(s) at some point(s).

At some point when I have energy back, I might write a polite “did you notice?” email along with a suggested fix (that all voices have the same text – to facilitate comparing intonation etc) and see what happens.

A quick note to Apple is probably a good suggestion. And since problem ins in their *text to speech” feature an update to the relevant text file or files would have an extremely small footprint for their test suites and bandwidth.

Sat, 2008-05-03 20:03 — 2126 (not verified) #permalink

That’s interesting indeed, though I think this entry will benefit from a bit of backstory. If I remember correctly, back in the early days of the Mac, there weren’t nearly so many voices available.

All of the voices we have today originated from the development of the PlainTalk suite. The speech synthesis component was a system extension called MacInTalk that first began shipping in System 7.1.2. However, MacInTalk was actually first demoed in 1984, when Steve Jobs unveiled the very first Macintosh. (You can watch a video of that moment, including the first-ever talking personal computer, on YouTube.)

Interestingly, back then, the only voice that existed was male, leading me to assume that while Steve Jobs was anthropomorphizing his technological invention, that he thought of it as a masculine entity. On the other hand, however, his company’s previous technical wonder was a computer named “Lisa”, which is obviously a female name and supposedly taken from the name of Steve Jobs’s own daughter (although I don’t think he has ever confirmed this).

So clearly Steve Jobs is comfortable anthropomorphizing computers either masculine or feminine entities.

Further, back in those days and still in some cases even today, it was shown that people responded more favorably to technology literally talking at them when the voice was a female-sounding one rather than a male-sounding one. This only makes me wonder who the subjects of these studies were (I suspect they were men), but it does explain why Apple would choose to ship MacInTalk with a number of female voices.

With the release of Mac OS X, Apple changed the name of MacInTalk to the plainer “Text-to-speech” feature, and also finally updated many of the voices. In doing so, however, they merely duplicated some of the female voices (i.e., Victoria still exists, but a more-natural clone called Vicki also exists), but they did actually create new male voices (i.e., Alex). In any event, I’m pretty sure that’s why the older female voices mostly say the same exact thing.

Today, there are an equal number of male and female voices users can select for Text-To-Speech, but the female voices were “born” way, way earlier than the commonly-used male voices.

So…yeah. Just a bit of trivia I guess. If anyone does write to Apple about this, I would love to see what they say—though I doubt they’d actually reply.

[I’d be curious too. I’m pretty sure all the different voices once had the same sentence so you could compare quality, and the same was true with Windows. Over time, though, I’m guessing it just morphed and since they morphed more male voices more recently… Thanks for the history lesson, May. Hope life’s treating you well down under. —fl]

Sun, 2008-05-04 00:39 — 2126 (not verified) #permalink

it was shown that people responded more favorably to technology literally talking at them when the voice was a female-sounding one rather than a male-sounding one.

That might explain a quirk of some of the recurring default statements—the male voices were talking about being inside the computer, while the female voices were talking about being the computer. (Best explanation I’d come up with before that was associating female computer voices with the computers voiced by Majel Barrett in Star Trek and male computer voices with HAL 9000 from 2001.)

[And actually I’m guessing they all started out saying the same thing but over time more work was done on male voices and so they wound up with newer samples too. (I sort of remember that male voices were harder to make sound “natural” so it’s not necessarily a problem that they did more work on those.) Thanks,Jfp. —fl]

Sun, 2008-05-04 21:22 — 2126 (not verified) #permalink

“the male voices were talking about being inside the computer, while the female voices were talking about being the computer.”

I’m not sure whether seeing the implications of that makes me a sexism-conscious feminist, or just lewd-minded. Not that they’re mutually exclusive – plenty of lewd-minded feminist geeks around here.

Sunflower

[Hey, that’s delightfully lewd. And almost certainly not Apple’s intention. (Have you seen their ads? No way either their idea of a PC guy or a Mac guy ever goes home let alone ever meets anybody else.) Thanks, Sunflower. —fl]

Post new comment

The content of this field is kept private and will not be shown publicly.
CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.