M$ Voice Recognition

[bewing]

A guy who sometimes pays me for freelance programming has asked me about an article he read. The article was indirectly about a new WinFX namespace that M$ has added to Vista, and the most recent version of Office, called System.Speech.
He wants me to find out info about programming it to his apps. Specifically -- how hard is their system to program with, for a simple interface?

So, I wanted to ask if any of you have dealt with it at all? Do you have any opinions, or other info beyond the propaganda that I can find on the M$ website? I don't currently have Vista, so I can't just try it and see. (And anyway, he hasn't budgeted any money for me to start experimental coding.)

[AndrewAPrice]

I've used MS SAPI. I'm not sure about the C# interface, but I got the SAPI (Speech API) working without any hassles at all. It was basically create device, send a string, send another string, and another, destroy device.

[bewing]

So you were going in the other direction? Sending text, and having it converted to voice for output? What interface did you use? C++?

[AJ]

I have tried both ways with the SAPI in VB6, but haven't tried since the days of .NET. I seem to recall that the voice recognition at the time wasn't wonderful, but then IIRC it was "untrained" speech recognition.

Interfacing with the API at the time was extremely easy and I should imagine with an IDE with good intellisense (MS's own Visual Studio.NET or SharpDevelop) and autocompletion it will be extremely hassle-free. Just check the distribution license for redistribution first (although if, as you say, it is buindled with Vista / Office there should be no trouble there).

Cheers,
Adam

[AndrewAPrice]

Oh, voice recognition. Sorry, I did voice synthesis.