MacSpeech Dictate v1.2

by David Marusek
Developer: MacSpeech
Developer URL: www.macspeech.com
Price: $199.00, includes headset

My config: MacBook, 2 GHz Intel Core Duo, 2 GB SDRAM
My review in a nutshell: If you want to dictate your speech directly into text, this is a good time to jump into voice recognition software. Dictate v1.2 is far from perfect, but it’s good enough to use now professionally, and it will only improve as it matures.
I am a professional science fiction writer with a peculiar handicap: something about a keyboard seems to rob me of my creativity. So I write out my stories and novels in longhand –two or three full drafts –before keystroking them into the word processor. No wonder then that I’ve been following speech recognition software for a while. Way back in 1993, I purchased a new Macintosh Quadra 840AV computer for my university office, the first Mac series to ship with voice commands (Speakable Items) and a built-in microphone. It didn’t work at all, no matter how loud I yelled at it. A few years later, I purchased v1.0 of MacSpeech iListen. It wasn’t much better. I decided to sit back and wait for the technology to mature enough to make it worthwhile to try again.

That time came last year when I had the manuscript for my second novel, Mind Over Ship, ready for keystroking. I had the opportunity to try Dragon Naturally Speaking v9 on my dentist’s office computer, and I was so impressed, I decided to buy a copy. DNS has no Mac version, and even though I already owned an Intel Mac, it would have cost about $500 to install it on my Mac (mostly to pay for Windows Vista). Instead, I loaded it on my daughter’s Dell laptop. I dictated the entire book into the Dell and copied it to my Mac for final editing. Dictating an entire book was a bumpy but successful operation, and I’m now a voice recognition believer. Which was why I was excited to hear that MacSpeech licensed the DNP voice rec engine and built a whole new native Mac application around it.

Dictate v1.2’s chief strength is its extraordinary accuracy. Almost no training is required. The headset included with the software is top-notch and comfortable to use. You can dictate directly into many software applications and navigate them by voice.

Dictate v1.2s chief weakness is its editing function. Granted, this function is completely new in this version, but it is so riddled with flaws that it is unusable. I’ll give just a couple of examples of this.

There are commands for moving the cursor forward or backward x number of words in your dictated text, and to the beginning and end of a document, but that’s about it. You can’t move back and forth by individual spaces, which one sometimes needs to do, or by a number of lines or paragraphs. There is no backspace command, which is very inconvenient. To select a word or phrase to edit, you speak it. Let’s say that the word “tiger” appears in your text three times and you want to edit its second occurrence. You say, “Select the word tiger.” If the software selects the first or third tiger, you’re out of luck. I could figure out no way of selecting the second tiger short of using my mouse. In DNS v9 (I know that v10 has been released, but I have no experience with it), you would simply repeat, “Select again,” until the desired tiger was selected, but Dictate lacks this option. If you run out of patience and select a word with your mouse, you have just ruined your editing session. The Dictate users manual emphasizes a “motto” for editing: If you’re talking, talk, and if you’re typing, type. This is by way of saying that Dictate uses some kind of mapping algorithm that only responds to voice commands. So the moment you use your keyboard or mouse to edit, you go off the internal map and the program loses its place –permanently, as far as I can tell. I consider this to be a fatal flaw and urge MacSpeech to fix it. There will be times when using the keyboard is the only way to do an edit, and if the internal mapping can’t handle it (as DNS seems able to do), you might as well forget about editing by voice altogether. Which is what I’ve pretty much done, for now at least. Dictate is so accurate that I’m satisfied getting the words in the word processor more or less correct as dictated the first time. From there I do all my editing with keyboard and mouse in Word.

Unfortunately, one needs to do a certain amount of editing by voice, especially at the beginning of the new document, in order to train the software to your writing style and further increase its accuracy. For example, let’s say you use a word that has homonyms (right, rite, write). If the software can’t figure out by context which spelling to use and uses the wrong one, and if you will be using that word frequently in the document, it behooves you to correct the error at the outset because Dictate is smart enough to learn your usage and keep from making the same error repeatedly. Therefore, editing is necessary, no matter how frustrating it is in the current version.

In conclusion, Dictate has a long way to go to match its rival, but I’m confident it will get there, and I’m in for the long haul now. Version 1.2 is already good enough to save me weeks of keystroking my next novel.