Raw Opus, Ogg and Wave Recorder

A javascript library originally forked from chris-rudmin/opus-recorder. This library handles encoding the output of Web Audio API nodes in

• Raw Opus
• Ogg Opus
• WAV format

The underlying implementation uses webassembly transpiled audio libraries integrated into the Javascript API.

Libraries Used

• Emscripten 2.0.8
• Libopus: v1.3.1 compiled with emscripten
• speexDSP: 1.2.0 compiled with emscripten

Required Files

The required files are in the dist folder. Unminified sources are in dist-unminified. Examples for recording, encoding, and decoding are in examples folder.

Usage

Constructor

The Recorder object is available in the global namespace and supports CommonJS and AMD imports.

import Recorder from 'raw-opus-recorder'
var rec = new Recorder({config-object});

Creates a recorder instance.

• config - An optional configuration object.

General Config options

• bufferLength - (optional) The length of the buffer that the scriptProcessorNode uses to capture the audio. Defaults to 4096.
• encoderPath - (optional) Path to desired worker script. Defaults to encoderWorker.min.js
• mediaTrackConstraints - (optional) Object to specify media track constraints. Defaults to true.
• monitorGain - (optional) Sets the gain of the monitoring output. Gain is an a-weighted value between 0 and 1. Defaults to 0
• numberOfChannels - (optional) The number of channels to record. 1 = mono, 2 = stereo. Defaults to 1. Maximum 2 channels are supported.
• recordingGain - (optional) Sets the gain of the recording input. Gain is an a-weighted value between 0 and 1. Defaults to 1
• sourceNode - (optional) An Instance of MediaStreamAudioSourceNode to use. If a sourceNode is provided, then closing the stream and audioContext will need to be managed by the implementation.

Config options for OPUS encoder

• encoderApplication - (optional) Supported values are: 2048 - Voice, 2049 - Full Band Audio, 2051 - Restricted Low Delay. Defaults to 2049.
• encoderBitRate - (optional) Target bitrate in bits/sec. The encoder selects an application-specific default when this is not specified. Example: (1s == 1000ms, frameSize == 20, encoderOutputMaxLength == 4000) -- (1000 / 20) * encoderOutputMaxLength === 200kbps desired. For perspective Spotify outputs between 96-160Kbps
• encoderComplexity - (optional) Value between 0 and 10 which determines latency and processing for encoding. 0 is fastest with lowest complexity. 10 is slowest with highest complexity. The encoder selects a default when this is not specified.
• encoderFrameSize - (optional) Specifies the frame size in ms used for encoding. Defaults to 20.
• encoderSampleRate - (optional) Specifies the sample rate to encode at. Defaults to 48000. Supported values are 8000, 12000, 16000, 24000 or 48000.
• maxFramesPerPage - (optional) Maximum number of frames to collect before generating an Ogg page. This can be used to lower the streaming latency. The lower the value the more overhead the ogg stream will incur. Defaults to 40.
• originalSampleRateOverride - (optional) Override the JS AudioContext recorded sample rate in the ogg opus 'input sample rate' field. Google Speech API requires this field to be 16000.
• resampleQuality - (optional) Value between 0 and 10 which determines latency and processing for resampling. 0 is fastest with lowest quality. 10 is slowest with highest quality. Defaults to 3.
• streamPages - (optional) dataAvailable event will fire after each encoded page. Defaults to false.
• rawOpus - (optional) output events will not be encapsulated in an Ogg frame. Defaults to false.
• encoderOutputMaxLength - (optional) sets the underlying max_data_bytes field in the call to opus_encode_float. defaults to 4000

Config options for WAV recorder

• wavBitDepth - (optional) Desired bit depth of the WAV file. Defaults to 16. Supported values are 8, 16, 24 and 32 bits per sample.

Instance Methods

rec.close()

close will close the audioContext, destroy the workers, disconnect the audio nodes and close the mic stream. A new Recorder instance will be required for additional recordings. if a sourceNode was provided in the initial config, then the implementation will need to close the audioContext and close the mic stream.

rec.pause([flush])

pause will keep the stream and monitoring alive, but will not be recording the buffers. If flush is true and streamPages is set, any pending encoded frames of data will be flushed, and it will return a promise that only resolves after the frames have been flushed to ondataavailable. Will call the onpause callback when paused. Subsequent calls to resume will add to the current recording.

rec.resume()

resume will resume the recording if paused. Will call the onresume callback when recording is resumed.

rec.setRecordingGain( gain )

setRecordingGain will set the volume on what will be passed to the recorder. Gain is an a-weighted value between 0 and 1.

rec.setMonitorGain( gain )

setMonitorGain will set the volume on what will be passed to the monitor. Monitor level does not affect the recording volume. Gain is an a-weighted value between 0 and 1.

rec.start()

start Begins a new recording. Returns a promise which resolves when recording is started. Will callback onstart when started. start needs to be initiated from a user action (click or touch) so that the audioContext can be resumed and the stream can have audio data.

rec.stop()

stop will cease capturing audio and disable the monitoring and mic input stream. Will request the recorded data and then terminate the worker once the final data has been published. Will call the onstop callback when stopped.

Instance Fields

rec.encodedSamplePosition

Reads the currently encoded sample position (the number of samples up to and including the most recent data provided to ondataavailable). For Opus, the encoded sample rate is always 48kHz, so a time position can be determined by dividing by 48000.

Static Methods

Recorder.isRecordingSupported()

Returns a truthy value indicating if the browser supports recording.

Static Properties

Recorder.version

The version of the library.

Callback Handlers

rec.ondataavailable( arrayBuffer )

A callback which returns an array buffer of audio data. If streamPages is true, this will be called with each page of encoded audio. If streamPages is false, this will be called when the recording is finished with the complete data. If rawOpus is true, this will be called with each encoded opus packet.

rec.onpause()

A callback which occurs when media recording is paused.

rec.onresume()

A callback which occurs when media recording resumes after being paused.

rec.onstart()

A callback which occurs when media recording starts.

rec.onstop()

A callback which occurs when media recording ends.

Getting started with webpack

• To use in a web-app built with webpack, be sure to load the worker files as a chunk. This can be done with file-loader plugin.

Add to your webpack.config.js before all other loaders.

module.exports = {
  module: {
    rules: [
      {
        test: /encoderWorker\.min\.js$/,
        use: [{ loader: 'file-loader' }]
      }
    ]
  }
};

Then get the encoderPath using an import

import Recorder from 'raw-opus-stream-recorder';
import encoderPath from 'raw-opus-stream-recorder/dist/encoderWorker.min.js';

const rec = new Recorder({ encoderPath });

Gotchas

• To be able to read the mic stream, the page must be served over https. Use ngrok for local development with https.
• All browsers require that rec.start() to be called from a user initiated event. In iOS and macOS Safari, the mic stream will be empty with no logged errors. In Chrome and Firefox the audioContext could be suspended.
• macOS and iOS Safari native opus playback is not yet supported
• The worker files need to be hosted on the same domain as the page recording as cross-domain workers and worklets are not supported.

Browser Support

Supported:

• Chrome v58
• Firefox v53
• Microsoft Edge v41
• Opera v44
• macOS Safari v11
• iOS Safari v11

Unsupported:

• IE 11 and below
• iOS 11 Chrome

Known Issues

• iOS 11.2.2 and iOS 11.2.5 are not working due to a regression in WebAssembly: https://bugs.webkit.org/show_bug.cgi?id=181781
• Firefox does not support sample rates above 48000Hz: https://bugzilla.mozilla.org/show_bug.cgi?id=1124981
• macOS Safari v11 does not support sample rates above 44100Hz

Building from sources

Prebuilt sources are included in the dist folder. However below are instructions if you want to build them yourself. Opus and speex are compiled without SIMD optimizations. Performace is significantly worse with SIMD optimizations enabled.

Mac: Install autotools using MacPorts

port install automake autoconf libtool pkgconfig

Windows: Install autotools using MSYS2

pacman -S make autoconf automake libtool pkgconfig

Install Node.js

Install EMScripten

Install npm dependencies:

npm install

checkout, compile and create the dist from sources:

npm run make

Running the unit tests:

npm test

Clean the dist folder:

make cleanDist

Clean the dist folder and the submodules:

make cleanAll