For the last few days I've been trying to get the Microsoft Speech Server SDK 1.1 to work on a PC that had a bunch of software installed, including both VS2003 and VS2005. What I was getting were error messages like "microphone wizard failed to initialize" and "The speech recognition stream was terminated" from the speech control panel.

This mystified me and I made the bad assumption that DotNet 2.0 was the problem based on previous issues that I'd had with Reporting Services 2003 vs 2005. So based on this assumption I set up a virtual PC with just XP sp2, VS2003 and the SDK -- lo and behold: it worked! This seemed to confirm my assumption, so I set about duplicating that configuration on the PC that I was using, by installing a second instance of XP that I planned to use just for speech server development.

I had XP installed and was just at the point where I needed to join the network domain when the boss wandered over to chat. I described the error message and went to demonstrate the error message to him by bringing up the speech control panel when I noticed an "add" button in the profile area of the speech control panel. I knew that the configure microphone button wouldn't work -- and it didn't -- I was able to demo this for the boss, but when I clicked on the add profile button it did work and I found that the speech engine started up and allowed itself to be trained.

Once I trained the speech engine using this add profile button then the configure microphone button started working as did the Speech SDK speech debugger. Odd that the default profile didn't work on that PC but did in the virtual PC...and I've just duplicated this behavior on a laptop that I have at home...error messages prior to adding a profile then once a new profile has been added and trained then the SDK seems to work fine...