Ask HN: Speaker independent voice recognition for auto-generated transcriptions?

Does anyone know of any code for doing speaker-independent voice recognition? I would love to be able to automatically transcribe various pieces of media.

I already know nothing is going to be even close to perfect. Is Sphinx still it? Is there anything out there that's great?

The code for YouTube's automatic transcription would be ideal. Someone tell me Google already open sourced it? </prayer>