From what I can tell, the audio is recorded by a Flash object and analyzed on a remote server, so calling it Javascript speech recognition is misleading.
Just a description of how to use some speech recognition APIs.<p>Would have been much more interesting if it was a speech recognition engine written on Node.js ;)