I am testing out a speech recognition using Dragon Dictate :-)
I strongly believe that the cloud vision API from Google is one of the most sophisticated Apis available on market. Google crowd sources a lot of information and is continuously evolving based on consumer feedback. This feedback loop allows Google to continuously improve its products and make them very good candidates for human interaction.
The speech API from Google is also very robust and time-tested. My experience with using Google speech provided by various browser applications has been very good. I currently use Dragon dictate from Nuance to type out this text. However, I think Google's API will be very flexible to create voice recognition applications.
These new services that have been opened by Google create wonderful opportunities for innovation for small and big developers alike. I really love to get my hands dirty with these new API's.