Microphone audio
Before the operating system asks for microphone permission, the game shows a just-in-time notice. On the web, audio is captured in temporary memory only while push-to-talk is active and sent over HTTPS to the configured speech-recognition service. In the Android app, the device speech-recognition service may process the transmission.
Talk to Tower does not retain that audio after transcription and does not use microphone audio to train a model. You may decline microphone access and use keyboard controls.
