WebRTC adapts to the audio signal. If we slice the audio clip into multiple shorter clips, then perform VAD on them in parallel, the result may not be as good.