Two ways to process voice, on device or on server. Device-based solutions either are very basic and just detect differences between words or need training data based on your voice or they need lots of processing power for more generalized voice recognition. So is your battery draining and phone is often hot because an app is keeping the mic on and keeping the phone from slowing the processor? Other option is to stream the data to the server. This would also increase battery usage as the phone can't sleep, but might not be as noticeable, but more evident would be your phone using a lot more bandwidth than is reasonable while you aren't actively using it.