If a single software program stands as a symbol of the Mac’s return to greatness, it’s the new ViaVoice from IBM. For years, Windows bigots have enjoyed pointing out that three different dictation programs were available (and insanely popular) for the PC. Speech-recognition software, went the refrain, was a classic example of the kind of software you just couldn’t get for the Mac. On a PC, you can say, “Minutes of the PTA meeting,” and your word processor types out, “Minutes of the PTA meeting.” On the Mac, all we had was the feeble PlainTalk, which can only open programs and pull down menus.
But those days are over. The Mac is cool again, and major software companies are bringing versions of their bestsellers to this new 25-million-person market. IBM’s ViaVoice is only the first of three Mac continuous-speech recognition programs to be released in the next few months. (Dragon Systems and MacSpeech are also readying Mac dictation software.)
ViaVoice is an astonishing bargain at $90, which includes an excellent Andrea noise-cancellation headset microphone that costs $40 when sold separately. (Only three years ago, I bought the now-discontinued PowerSecretary, a Mac speech-recognition program that required you to separate each word with a pause. The price: $2,500. Is it possible to weep and dictate at the same time?)
The headset even includes a comfortable ear speaker, which lets you listen to the Mac read your text back without bothering other people in the room. And if there was any doubt that IBM was serious about creating a genuine Macintosh program, the headset even includes five sets of translucent plastic accent panels — one for each iMac/iBook fruit color.
But then there’s the software. It’s saddled with one whopping limitation: you can dictate only into its own little SimpleText-type word processor. From there, voice commands can almost instantly paste what you’ve dictated into your choice of five programs: America Online, Word 98, AppleWorks, Netscape Messenger, or Outlook Express.
In about two minutes, owners of programs like QuicKeys and OneClick can make macros that transfer your dictated text to any program. Still, anyone who’s ever experienced the thrill of speech-editing directly in Word (using a program like Dragon NaturallySpeaking for Windows) will ache for the same kind of directness on the Mac.
Otherwise, the news is mostly good. The package is beautifully designed, shining with an attention to Macintosh-specific detail at every step. For example, the Setup Assistant doesn’t just tell you to plug the headset cable into the microphone jack on your Mac — it actually shows a photograph of your Mac model’s back panel, using some clever internal scheme to figure out what model Mac you’ve got. (You’ll be able to use ViaVoice with your iBook laptop as soon as the Griffin iMic adapter ships in late January, according to Griffin Technology.)
You’re then asked to acquaint the software with your particular voice by reading several screens of text out loud. On a Power Mac G3/300, reading the shortest of these excerpts (from, for example, “Treasure Island”) takes about 15 minutes; on an iMac/400, it takes about 10 minutes. The software then goes on sabbatical for about 30 minutes, analyzing the sound files it has recorded.
Then the fun begins. From the permanent ViaVoice icon on your menu bar, you launch the SpeakPad application (the ViaVoice version of SimpleText). You put on the headset, start talking — and absolutely nothing happens. Then, suddenly, at the first pause in your speaking, the transcription of your words appears on the screen.
It Works!Having lived and breathed Windows dictation software for three years — yes, even wrist-disabled Mac diehards like me must use Windows when there’s no alternative — I can say that in the scheme of dictation software, the ViaVoice accuracy is excellent. On your first afternoon, you’ll get about 95 percent accuracy, which means that you’ll have to correct one out of 20 words. The accuracy creeps upward as you keep using the program and making corrections. An optional feature lets you import a bunch of writing you’ve already done, so that ViaVoice can spot oddball words that crop up in your line of work — say, “twelvish” and “Clintonesque” — which further improves accuracy.
Fortunately, you can make corrections entirely by voice, saying, for example, “correct ‘mode import.’ A numbered list of alternative transcriptions appears immediately. You then speak the number (“pick seven”) of the transcription you intended, such as modem port. The correction appears instantaneously in your dictated text, even if it’s several sentences back.
Unfortunately, you then experience the single truly brain-dead design flaw of ViaVoice: after you make a correction, your insertion point blinks at the spot in the text where you made the correction. Inevitably, that’s several phrases or sentences back. You’re forced to use the mouse to click at the end of your document in order to resume dictation. (After you make a correction in other programs, like Dragon NaturallySpeaking, your cursor gets dumped where you stopped dictating, which makes a thousand times more sense.)
Other glitches: The program always puts two spaces after a period–you have no vote on the matter. ViaVoice requires that virtual memory be turned on and set to at least 10MB higher than your actual installed RAM. Making a capitalized word is way too unwieldy: you have to say “Capitalize this” before the word, which is about three syllables too long. (In Naturally Speaking, you just say “cap.”) And I couldn’t get the New User command to work.
The Speed of ThoughtIs ViaVoice dictation faster than typing? It certainly isn’t as fast as the best Windows dictation software, and plods on 233-MHz Macs (the minimum suggested platform). On an iMac DV, you wait about one second after each utterance for the text to appear on the screen. When assessing the words-per-minute count, you must also factor in the time it takes you to make corrections, which decreases as the months go by.
But to answer the question: On the first day of using the program, most touch typists will probably average the same speed they get when typing. As you and the program get to know each other, the speed of dictating pulls ahead. In the end, though, what you’ll discover is that the speed of ViaVoice is mostly irrelevant. The bottleneck, when you write by dictating, isn’t the computer — it’s your brain. As you sit there, mentally composing each sentence before you speak it, the software winds up waiting for you.
Fortunately, the transition from keyboard to microphone doesn’t take as long as you might expect. If you’re a person who can’t, or doesn’t like to, use the keyboard, ViaVoice is a breakthrough on the Macintosh. To be sure, the fact that you’re confined to dictating in a proprietary word processor brands ViaVoice as a 1.0 product. But at least IBM put its emphasis in the right areas: accuracy and a fanatical devotion to the Macintosh way. The program is dirt cheap, and comes with a 30-day guarantee. Anyone who types anything longer than URLs each day should investigate this remarkable program, if only to have been acquainted with the thrill of real Mac dictation software when ViaVoice’s more powerful rivals arrive in the spring.
David Pogue is the coauthor, with Adam Engst, of Crossing Platforms, a two-way Mac-Windows dictionary (O’Reilly, 1999).