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.
Through flash it streams audio live to a red5 server where sphinx is used for decoding. Open sourcing the red5 app soon so hopefully people can get use out of it whether embedded in a webpage or as a flash application(such as a flash game). It will be easier with html5.
4 comments
[ 3.7 ms ] story [ 20.7 ms ] threadWould have been much more interesting if it was a speech recognition engine written on Node.js ;)
Through flash it streams audio live to a red5 server where sphinx is used for decoding. Open sourcing the red5 app soon so hopefully people can get use out of it whether embedded in a webpage or as a flash application(such as a flash game). It will be easier with html5.