Are the readings done via mic + real in the flesh speaker? Or are you using tts?