The recently announced smartwatches based on Android Wear extensively rely on voice recognition for user interaction.
Is the voice recognition performed on-device, on the paired smartphone, or in Google's cloud? Or is it a mix of those three, e.g. online with an offline fallback?