[Simucal]

Android has offline speech recognition (introduced in Jelly Bean). From my limited testing it works really well. So, I don't think an external server is as required as some people say.

It works insanely fast, transcribing what I say in near real time which feels like black magic compared to Siri on my iPhone that has to record an audio clip in its entirety, send it up to their servers, process it, then send a response back.

[jonknee]

Glass could use the Android handset as a remote server--it shouldn't matter if the Android crunches voice on its own or with a data connection, all that matters is Glass getting a reply from the API call.