Wednesday 28th October, 1998
Speech recognition software isn’t new to the Mac platform. Those with a long memory may recall VoiceNavigator, an external piece of signal-processing hardware strapped to your IIci that could recognise predefined commands for working with windows, menu items, and buttons. Those who managed to get it to work found their life expectancy halved, as office colleagues were driven mad by incessant ‘close window’ commands.
VoicePower Pro, on the other hand, is the real thing. It works by listening to discrete words, and matching them against entries in its phoneme-to-English dictionary. In theory, what you say into the microphone ends up on the screen.
The package, a rebadged and updated version of PowerSecretary, includes software and hardware in the form of a matched pair of microphone and headset and a PlainTalk pre-amp. Installation should be straightforward: run the installer from the CD, connect the pre-amp to the microphone-in socket on the back of your Mac, and the headset to both the pre-amp and the Mac’s headphone-out socket, restart your Mac, open the VoicePower Pro application and begin training it to your voice.
That’s the theory. The problem is there are two versions of the VoicePower Pro package available in the UK. One Stop Direct provides a US version of the software packaged with a low-cost microphone and pre-amp, which only works with higher-end models directly: Performa (5000 and 6000 series) and PowerBook 1400 and 5300 owners require an upgraded microphone and pre-amp costing £60 (£700 inc VAT) as the pre-amp supplied may damage 1400 and 5300 PowerBooks. As it’s a US version, your words will be translated into US-speak: color for colour, honor for honour, gray for grey, and so on. A UK dictionary is an added option - again costing £60 (£70 inc VAT), or as the documentation helpfully suggests you can work through and correct the 3000 US spellings as you go. Having to pay for these two options is a bit of a swizz.
The second version of the package, available from VoiceWorks, includes the UK dictionary and the upgraded pre-amp/ microphone pair for £250 (£294 inc VAT), saving you £20 if you’re a Performa owner who wants to spell words in UK English.
Having loaded the software and installed the correct hardware for your machine, using it is fairly straightforward.
When you first open the VoicePower Pro application, you’re asked to read a short passage into the microphone, one highlighted word at a time. From this short passage, the software decides which of the two ‘standard’ voices (A and B) are the closest match to your voice. It then loads this voice and you move on to the training proper. You read a number of short paragraphs, one word at a time, and the application figures out and corrects the differences between your voice and the chosen standard voice, saving these differences as it goes. At the start you may find yourself re-reading words two or three times before being allowed to proceed to the next word. In some cases it’ll come across a word it really doesn’t understand, and you’ll be presented with a dialog box that concentrates on just that word.
After five or so training paragraphs, it saves its changes and you’re offered the chance to rest your voice. Even during the training period you’ll soon begin to notice an improvement in the speech recognition. One criticism of the training is its obvious US bias - it seems strange to be teaching an application how you pronounce words repeating US Indian place names, such as Chicago, Milwaukee, Illinois, and so forth. Having finished training, you’re advised to make a backup of your individual voice and vocabulary files - a good idea as you probably don’t want to repeat the training too often.
In use, you need to run the VoicePower application alongside any application you wish to use voice recognition in, which brings up the subject of memory. Memory-wise speech recognition is very hungry, and not at all partial to virtual memory. In plain English this means that since the application itself requires roughly 14Mb, running under Mac OS 8.1, you’ll need a recommended minimum of 40Mb of real memory.
Once your chosen application and VoicePower Pro are running together, the program appears as a palette in the other application. You can switch the microphone on and off by clicking an icon in the palette, or using the microphone hot-key (command-option-m). With the microphone on, you simply speak each word separately, and VoicePower Pro will translate each word you say into text which it adds to a scrolling list of words in the palette, and inserts at the cursor in the main application.
By default, VoicePower Pro learns as it goes along, so each word it correctly identifies reinforces its understanding of the way you speak. This means it’s important to correct mistakes as you go. To correct a word, you simply say ‘correct word’. A mini-palette pops up which offers up to nine alternatives which you can choose directly by saying ‘choose one’ or whatever number corresponds to the word you want. If the correct option isn’t shown, you can spell out the word using the international pilots' alphabet (alpha, bravo, charlie, and so on).
As you spell, the options in the list change to reflect the possible available options, and at any point you can choose an option. Having spelt your word and/or chosen one of the options, the palette disappears, and the new word substitutes the incorrect one.
The real value of VoicePower Pro is in the ability to enter text and control items on the screen without using a mouse or keyboard. The fact it only recognises discrete words means it’s not as fast as a typewriter. But for applications where your hands are needed elsewhere, or you wish to avoid RSI doing simple repetitive tasks with the mouse (VoicePower Pro includes a very useful AppleScript addition, with events that let you mimic keyboard work directly) then it’s ideal. The more you use it, and the faster the machine you’re using it on the better - it’s very slow on a 1400, bearable on a 6400 and usable on a G3.
Software and hardware combination that enables you to control the Mac with your voice.
|Get Further Info|
|Pros||Does what it says on the tin|
|Cons||Only handles discrete words ˜ Options for US version are confusing ˜ Power Mac only ˜ Can be slow|
|Originally Published||MacUser, Volume: 14, Issue: 20, p24|