Initial reactions to ViaVoice

We've started to receive some feedback regarding IBM's recently released ViaVoice software:

Frank McCulley was disappointed with it: "I went through all the preliminary training steps and it still garbles my words more than I do. For instance, after a whole day of working with this I just said the words "via voice" into the thing 10 times. Each time I said it I got a totally different response on my screen."

Raul Gutierrez was more upbeat: "After finishing with the setup assistant, I simply started the IBM SpeakPad program. Recognition is pretty good... in fact better than I expected and better than I have seen with similar software running on the PC. If the software is to be believed, the program keeps on learning and getting better. After a few hours with it, I'll say they seem to be right. I completed this email without ever touching the keyboard in about the same time it would take to type."

Regarding the initial setup procedure, he adds: "Setup was problem free. The Setup Assistant is easy to use although you should set aside time to run it. Reading all the learning text takes a minimum of 20 minutes... more if you want to teach it more. After reading the text, the program analyzes it. You should not try to do anything else while this is happening because the program takes over both the CPU and the hard drive. In fact the hard drive spun harder than I've ever heard it spin before. The program predicted 45 minutes of processing - on my 400MHz G3 it took about 15 minutes."

Norm Birtwistle writes of feature he would like added: "According to information on IBM's web site, ViaVoice requires that you dictate into their vehicle and then transfer it into your email or word processor application. (Apparently, Dragon Dictate will be the same when it eventually arrives.) It would be more convenient if you could dictate directly into your word processor etc." Maybe version 2.0 will do that.

Update: Matt Deatherage replies: "Not likely. What you're asking for is essentially full-blown continuous speech recognition in the _background_, resolved to any application by posting the speech as key-down events. Maybe someday with very fast processors and preemptive multitasking, but not with today's Mac OS and cooperative multitasking model."