Is this using IBM Watson for STT? Are you assuming the costs?

Or does it use some windows dictation api?