Export and the player

When a recording finishes, onComplete gives you a file and a compact summary of it:

onComplete={(e) => {
  e.uri;           // 'file:///.../wfr_....m4a'
  e.durationMs;    // 7340
  e.format;        // 'm4a'
  e.mimeType;      // 'audio/mp4'
  e.sizeBytes;     // 58123
  e.sampleRate;    // 44100
  e.channels;      // 1
  e.samples;       // 64 numbers in [0, 1]
  e.peakAmplitude; // 0.93
}}

Formats

Set the format with output.format. The file extension and mimeType depend on the platform:

format	iOS file	iOS mimeType	Android file	Android mimeType
m4a (default)	AAC in .m4a	audio/mp4	AAC in .m4a	audio/mp4
aac	AAC in .m4a	audio/aac	AAC in .m4a	audio/mp4
wav	16-bit PCM .wav	audio/wav	16-bit PCM .wav	audio/wav
opus	Opus in .caf	audio/opus	Opus in .ogg (API 29+)	audio/ogg

• m4a is the safest choice for sharing and playback everywhere.
• aac records the same audio as m4a. Only the reported MIME type differs on iOS.
• wav is lossless and large. It is the only format that supports the PCM stream.
• opus gives the smallest files, but the containers differ: .caf on iOS is not widely playable outside Apple platforms. On Android below API 29 the engine records AAC instead and fires onError with code 'format-unsupported'; the file keeps the .ogg name and audio/ogg type even though it contains AAC.

On Android output.bitrate is clamped by output.quality: 32 to 64 kbps for low, 64 to 128 kbps for medium and 96 to 256 kbps for high. On iOS the bitrate and quality are passed to the encoder as is.

The 64-bar export

samples is a ready-made waveform for a chat bubble, so the receiving side does not have to decode the audio. It is computed natively from the meter readings of the whole session:

1. The readings (one per meter tick, meterUpdatesPerSecond per second) are split into 64 equal groups in time.
2. Each group becomes the root mean square of its readings.
3. All 64 values are divided by the largest one, so the loudest bar is 1.

Things to know:

• The values are relative. A quiet recording and a loud one both reach 1. Use peakAmplitude if you need the absolute level.
• A recording with fewer than 64 readings (under about two seconds at the default 30 per second) leaves some groups empty, and those bars are 0.
• Readings come from the meter, so they follow the platform's meter mapping (see onMeter). The same sound gives slightly different shapes on iOS and Android.
• Very long sessions keep at most 16,384 readings; older ones are merged in pairs, keeping the louder value.

Keep the file

The file in onComplete.uri usually lives in the app's cache directory, and it still belongs to the recorder session:

• cancel(), also after stop(), deletes the session's segment files.
• Removing the view from the screen does the same: Android cleans up when the view is dropped, iOS when it is recycled.
• For a recording without a preview and continue step, the segment file is the delivered file.

So in onComplete, move or copy the file to a permanent location (for example the documents directory) or upload it before you cancel, start a different screen or unmount the recorder. Starting a new recording with start() does not delete the previous file.

Play it with react-native-waveform-player

react-native-waveform-player is the playback half: the same pill, bars and timer, with play, pause, scrubbing and speed control. Pass the URI and the 64 bars, and the player skips decoding the file:

import { AudioWaveformView } from 'react-native-waveform-player';

<AudioWaveformView
  source={{ uri: note.uri }}
  samples={note.samples}
  style={{ height: 56 }}
/>

Both libraries are independent and have no shared dependencies. If you send voice notes to other users, send samples along with the file (64 numbers are a few hundred bytes of JSON), so the receiving device can draw the bubble right away.