Node.js implementation of Web audio API

Node Web Audio API

This library implements the web audio API specification on node.js.

And this is not even alpha. Use this library only if you're the adventurous kind.

  • AudioContext(partially)
  • AudioParam (almost there)
  • AudioBufferSourceNode
  • ScriptProcessorNode
  • GainNode
  • OscillatorNode (coming soon)
  • DelayNode (coming soon)
npm install web-audio-api

Get ready, this is going to blow up your mind :

npm install
gulp default
node test/manual-testing/AudioContext-sound-output.js

By default, node-web-audio-api doesn't play back the sound it generates. In fact, an AudioContext has no default output, and you need to give it a writable node stream to which it can write raw PCM audio. After creating an AudioContext, set its output stream like this : audioContext.outStream = writableStream.

This is probably the simplest way to play back audio. Install node-speaker with npm install speaker, then do something like this :

var AudioContext = require('web-audio-api').AudioContext
  , context = new AudioContext
  , Speaker = require('speaker')
context.outStream = new Speaker({
  channels: context.format.numberOfChannels,
  bitDepth: context.format.bitDepth,
  sampleRate: context.sampleRate
// Create some audio nodes here to make some noise ... 

Linux users can play back sound from node-web-audio-api by piping its output to aplay. For this, simply send the generated sound straight to stdout like this :

var AudioContext = require('web-audio-api').AudioContext
  , context = new AudioContext
context.outStream = process.stdout
// Create some audio nodes here to make some noise ... 

Then start your script, piping it to aplay like so :

node myScript.js | aplay -f cd

icecast is a open-source streaming server. It works great, and is very easy to setup. icecast accepts connections from different source clients which provide the sound to encode and stream. ices is a client for icecast which accepts raw PCM audio from its standard input, and you can send sound from node-web-audio-api to ices (which will send it to icecast) by simply doing :

var spawn = require('child_process').spawn
  , AudioContext = require('web-audio-api').AudioContext
  , context = new AudioContext()
var ices = spawn('ices', ['ices.xml'])
context.outStream = ices.stdin

A live example is available on Sébastien's website

Gibber is a great audiovisual live coding environment for the browser made by Charlie Roberts. For audio, it uses Web Audio API, so you can run it on node-web-audio-api. First install gibber with npm :

npm install gibber.audio.lib

Then to you can run the following test to see that everything works:

npm test gibber.audio.lib

Each time you create an AudioNode (like for instance an AudioBufferSourceNode or a GainNode), it inherits from DspObject which is in charge of two things:

  • register schedule events with _schedule
  • compute the appropriate digital signal processing with _tick

Each time you connect an AudioNode using source.connect(destination, output, input) it connects the relevant AudioOutput instances of source node the the relevant AudioInput instance of the destination node.

To instantiate all of these AudioNode, you needed an overall AudioContext instance. This latter has a destination property (where the sound will flow out), instance of AudioDestinationNode, which inherits from AudioNode. The AudioContext instance keeps track of connections to the destination. When that happens, it triggers the audio loop, calling _tick infinitely on the destination, which will itself call _tick on its input ... and so forth go up on the whole audio graph.

Right now everything runs in one process, so if you set a break point in your code, there's going to be a lot of buffer underflows, and you won't be able to debug anything.

One trick is to kill the AudioContext right before the break point, like this:


that way the audio loop is stopped, and you can inspect your objects in peace.

Tests are written with mocha. To run them, install mocha with :

npm install -g mocha

And in the root folder run :

npm test

To test the sound output, we need to install node-speaker (in addition of all the other dependencies), and build the library :

npm install
npm install speaker
gulp default
node test/manual-testing/AudioContext-sound-output.js

To test AudioParam against AudioParam implemented in a browser, open test/manual-testing/AudioParam-browser-plots.html in that browser.

    61  Sébastien Piquemal
    16  ouhouhsami
     4  John Wnek
     2  anprogrammer
     1  Andrew Petersen
  • removed node-speaker and mathjs dependencies
  • now use aurora installed from npm instead of distributing a built version of it.
  • refactored to ES6
  • AudioNode and AudioContext bug fixes
  • audioports : bug fixes
  • audioports : implemented channelInterpretation 'speakers'
  • AudioContext : added support for mp3 to decodeAudioData
  • AudioBufferSourceNode : handler onended implemented
  • AudioContext : method decodeAudioData, support only for wav
  • ScriptProcessorNode

  • AudioBufferSourceNode

    • node is killed once it has finished playing
    • subsequent calls to start have no effect
  • AudioContext : method collectNodes

  • audioports : bug fixes

  • AudioContext (partial implementation)
  • AudioParam (missing unschedule)
  • AudioBufferSourceNode (missing onended)
  • GainNode