Hearing voices: How does Apple's VoiceOver feature perform?
Apple's latest iPod Shuffle is tiny. So small in fact that there isn't any room on the actual player for the control buttons – these have been moved to the headphone cord. But some complain that tiny players like this lack an essential ingredient: the ability to know what song you're listening to.
Apple's answer? VoiceOver. It's an optional feature that you can turn on when using your third generation iPod Shuffle.
By pressing and holding the small control button on the Shuffle's audio cord during a song, you can summon up the song's title and artist information, spoken in a silky-smooth voice.
If you're a Mac user, you're probably used to the quality of Apple's synthetic speech – it's pretty good. But I was curious how VoiceOver would perform in the accuracy dept. So I loaded up the Shuffle with a full 4GB of music and started going through the tracks, one by one, listening carefully for any glitches.
The verdict: VoiceOver is nearly – nearly – flawless. I threw it some real curve balls, in anticipation of a fumble but for the most part, the voice kept up.
Some unexpected successes:
E=MC2 by Big Audio Dynamite. Given that the "2" in the equation isn't superscript (you can't do superscript in song info) I would have forgiven VoiceOver had it said "E equals em see two" but it recognized the equation and pronounced the exponent correctly as "squared".
Stephen Colbert. I realize there are probably very few people with this last name who actually pronounce the 't' at the end, but with synthesized voice you just never know. No problems here, VoiceOver said "Colb-air".
The VoiceOver feature isn't dependent on just the Shuffle for its accuracy. The algorithm works in tandem with iTunes which performs the analysis on the song information and decides how VoiceOver will handle each track. Apple claims that it can even tell if your music originates in another part of the world, and will choose an appropriate voice for the song. E.g. iTunes recognizes the song “Ya Viene el Sol” by Ozomatli as a Spanish song and thus speaks the information in a Spanish voice.
But the system isn't flawless. For some reason, iTunes decided that Roxette's Little Girl and Peter Schilling's Major Tomweren't English tracks and tried to read the titles in some other language resulting in something sounding like "Leeall Yerl" and "Mahyoar Toam". Thankfully, if iTunes ever guesses wrong as in these cases, you can force it to adopt the language of your choice in the options tab in the track info dialog box.
Some other oddities were:
Natalie Imbruglia. Her last name has a silent 'g', but VoiceOver decided to include it.
R.E.M.'s Crushed With Eyeliner sounded like "Crushed with eye-linner"
Sammy Hagarlost his long "a" in the last name, becoming "Heh-GAR"
Tom Petty's Two Gunslingerssounded a lot like "Two Gun-sleevers"
And lastly and perhaps most jaringly, AC/DC became "Ah See Dee See". This is one of those times an "a" really needs to sound like "Ay".
To sample what VoiceOver sounds like, check out Marc's fun video overview of the latest Shuffle, or head on over to Apple's site for a demo of the international voices.
After four to six weeks, they return for additional training. ,
LikeLike