In case you're wondering about the accuracy of the software, examine the following paragraph:

One has to use care occasionally when speaking phrases program recognizes as commands. The developers have carefully chosen command phrases which do not appear frequently in regular speech. If, however, one wishes to document such a phrase or otherwise to enter it literally as text, it is necessary to separate the words with long pauses so that the phrase as a whole is not recognized by the software. For example, I am here dictating the phrase "what can I say", which normally displays a list of available verbal commands. I did this by speaking the words one by one rather than stringing them together.

This was entered using ViaVoice and the only correction was to remove extra spaces around the quotation marks. I got lucky with this test paragraph -- usually there are a couple of mistakes in a paragraph of text. I did find, though, that ViaVoice got better at recognizing my speech once I started correcting it on words that I use most often. Note in the preceding paragraph that there is a word "the" missing from the first sentence. This is typical of the types of mistakes that are encountered.

Compared to other speech recognition systems I've seen, the accuracy of ViaVoice is superb and I can only assume it will improve with training of both the software and the user. There is definitely a knack to getting accurate results with text-to-speech conversion, and I can't really describe it. Suffice it to say that you will learn what works and what doesn't, and you will become able to use the software faster and more naturally over time. You can also re-run the enrollment program, choosing different text samples than the one you used first, to improve the accuracy of the speech recognition engine.

Next: My mind is going. I can feel it! »