Professional version brings voice recognition, navigation, and control to Microsoft Office. by Stan Miastkowski, special to PC World September 14, 1998, 12:44 p.m. PT Repeat after me: Competition is good. A certain large software company in Redmond may not agree, but the proof certainly shows in voice recognition software. Over the past few months, I've been working extensively with voice software. And as the four major companies in the genre--Dragon Systems, IBM, Lernout & Hauspie, and Philips--slug it out, a leapfrog effect has started, with each company coming out with better software. And prices keep falling, too. With its latest products, L&H has taken an approach similar to IBM ViaVoice, offering three versions of Voice Xpress, each designed for different kinds of users. Voice Xpress Standard ($49.99) offers basic voice recognition, Voice Xpress Advanced ($79.99) is designed for use with Microsoft Word, and Voice Xpress Professional ($149.99) offers full-fledged support for all Microsoft Office applications. I tested a prerelease version of Voice Xpress Professional, and it goes way beyond its predecessor with a raft of useful features. And speech-to-text conversion is much improved, to boot. Voice Xpress Professional has a new "modeless" design, similar to its competitors' products, that lets you switch between dictation, editing, navigation, and desktop control using voice alone. When you're dictating, you just talk continuously. Commands ("move to beginning of document," for example) are detected by your pauses before and after, and navigation and desktop control ("Open Microsoft Excel" or "Switch to PowerPoint") respond to specific phrases. It sounds complicated, and you do need to become familiar with the various key phrases, but Lernout and Hauspie has done an excellent job of creating a powerful and integrated voice package tuned to Microsoft Office users. As with all voice packages, you'll need to do some preliminary work, spending 15-30 minutes training Voice Xpress Professional to your voice. (Different users can create individual profiles.) Basic voice recognition is very good, and it gets better over time, as the package continues to tune itself to your idiosyncrasies. Of course, you'll also have to train it to recognize words that aren't in its vocabulary. No voice package understands "Miastkowski" the first time I say it. L&H says you need a minimum of 48MB of RAM and a Pentium-200, but the faster your CPU, the better the performance. Even the PII-333 system with 64 MB of RAM I used slowed down occasionally. Voice recognition isn't for everyone; speaking is vastly different from typing, and requires some getting used to. But if your fingers are getting tired, Voice Xpress Professional does the job and is close to the current state of the art--until the next version, that is.