Manual and Real-Time modes allow for natural voice interaction that detects when you finish speaking [cite: 6.2, user_prompt].
Custom audio filtering handles loud factory floors by removing background rumble and normalizing volume [cite: 6.10].