Silence detection

The recorder can tell you when the input has been quiet for a while, for example to stop a voice memo once the user stops talking.

<WaveformRecorderView
  ref={ref}
  style={{ height: 56 }}
  silenceThresholdDb={-45}
  silenceTimeoutMs={2000}
  autoStopOnSilence
  onSilenceDetected={(e) => console.log(`Quiet for ${e.durationMs} ms`)}
  onComplete={(e) => send(e.uri, e.samples)}
/>

Prop	Default	Meaning
silenceThresholdDb	-160	Level in dBFS below which a meter reading counts as quiet.
silenceTimeoutMs	0	How long the input must stay quiet. 0 turns detection off.
autoStopOnSilence	false	Call stop() automatically when silence is detected.

The defaults never fire: -160 dBFS is below any real reading and a timeout of 0 disables the check. Set both the threshold and the timeout.

How it is measured

Silence detection runs on every meter tick while recording, meterUpdatesPerSecond times per second:

1. If the tick's db is at or above silenceThresholdDb, it counts as sound and the quiet timer resets.
2. Otherwise the quiet stretch continues. The first quiet tick after sound starts the timer.
3. Once the stretch reaches silenceTimeoutMs, onSilenceDetected fires with how long it has lasted, and with autoStopOnSilence the recording stops and onComplete fires.
4. It fires once per quiet stretch. The next tick at or above the threshold re-arms it.

Each tick is compared on its own. There is no averaging over several ticks, so a single loud tick, such as a click, resets the timer.

The timer uses wall-clock time and is not reset by pause(). If the user pauses in the middle of a quiet stretch, the paused time counts too, and the event can fire on the first quiet tick after resume().

Choosing a threshold

The db value is the same one onMeter reports, and it is measured differently per platform:

Platform	What db is
iOS	Average power over the last tick (AVAudioRecorder.averagePower).
Android	Peak level over the last tick, converted to dBFS.

Peaks are higher than averages, so the same room reads louder on Android. Start around -45 to -50 dBFS, log onMeter on a real device in a quiet room and while talking, and pick a value between the two, per platform if needed:

silenceThresholdDb={Platform.OS === 'ios' ? -50 : -40}

You can also watch the numbers in the browser demo, which computes both mappings from your microphone.