virtual-audio-graph

virtual-audio-graph is a library for declaratively manipulating the Web Audio API that is inspired by React and virtual-dom. This guide aims to introduce it step-by-step to help you get up and running as quickly as possible.

Importing

First let's import everything we're going to need for these examples:

import createVirtualAudioGraph, {
  createNode,
  createWorkletNode,
  delay,
  gain,
  oscillator,
  stereoPanner,
} from 'virtual-audio-graph'

Creating A virtual-audio-graph Instance

With No Configuration

Next let's create our virtual-audio-graph instance:

const virtualAudioGraph = createVirtualAudioGraph()

With Configuration (Optional)

createVirtualAudioGraph optionally takes a configuration object that let's you specify the AudioContext instance to use and the output the audioGraph should be connected to (any valid AudioNode destination). If no audioContext is specified then a new one is automatically created and if no output is specified then it defaults to the audioContext destination.

Here's what it looks like to pass your own configuration to virtual-audio-graph:

const audioContext = new AudioContext()

const virtualAudioGraph = createVirtualAudioGraph({
  audioContext,
  output: audioContext.destination,
})
Note that the number of instances of AudioContext that can be created is limited so if you already have one it may be best to provide it here.

virtual-audio-graph Instance Public Interface

Here is everything we can do with our virtual-audio-graph instance:

Rendering Our First Graph

virtualAudioGraph.update takes an object that represents the underlying audio graph. Each key is used as the id for a particular node and each value specifies all the attributes we need to know about that node.

In this example we are creating a gain node with id 0 and an oscillator node with id 1.

For each virtual audio node factory the first argument configures how to connect the node and the second is optional and configures the parameters to pass to the node:

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: gain('output', { gain: 0.5 }),
  1: oscillator(0, { stopTime: currentTime + 1 }),
})

We can see that the gain node is being connected to "output" which is a special string reserved for the virtual-audio-graph instance output (specified when our virtual-audio-graph instance is created) and has its gain value set to 0.5. We can also see that the oscillator node is being connnected to the gain node (id 0) and has its stopTime set to 1 second from the current time.

Rendering An Empty Graph

All the demos in this guide have a stop button that is implemented like this:

virtualAudioGraph.update({})

Rendering an empty graph removes all the nodes and brings us back to our initial state. The power of virtual-audio-graph is that all the Web Audio API imperative code is handled for us and we don't have to worry about it.

Another Basic Graph

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: gain('output', { gain: 0.2 }),
  1: oscillator(0, {
    frequency: 440,
    stopTime: currentTime + 2.5,
    type: 'sawtooth',
  }),
  2: oscillator(0, {
    detune: 4,
    frequency: 554.365,
    startTime: currentTime + 0.5,
    stopTime: currentTime + 2.5,
    type: 'square',
  }),
  3: oscillator(0, {
    detune: -2,
    frequency: 660,
    startTime: currentTime + 1,
    stopTime: currentTime + 2.5,
    type: 'triangle',
  }),
})

We've now connected 3 oscillators to our gain node and provided them with a few more parameters and virtual-audio-graph updates all the underlying audio nodes for us.

Specifying Multiple Connections & Connecting To AudioParams

The output parameter of the node factory functions is a lot more versatile than just specifying a single connection. You can use it to specify connections to multiple nodes and/or multiple connections to AudioParams of those nodes.

If you wish to make more than 1 connection just use an array with each element as either a node key, the special string "output" or an object with a key property and a destination property that specifies the node and AudioParam to connect to:

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: gain('output', { gain: 0.2 }),
  1: oscillator(0, { stopTime: currentTime + 3 }),
  2: gain({ destination: 'frequency', key: '1' }, { gain: 350 }),
  3: oscillator([2, 'output'], { frequency: 1, type: 'triangle' }),
})

In the above example we have connected:

In this way you can start to specify any sort of graph that the Web Audio API allows

AudioParam Methods

If you're familiar with the Web Audio API you will know that AudioParams have methods as well as values and virtual-audio-graph allows you to update these too. Just specify an array where the first element is the method name as a string and the remaining elements are the arguments for that method. If scheduling multiple values specify an array of these arrays. (See here for more info on AudioParam methods).

Here's how to use setValueAtTime:

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: gain('output', { gain: 0.5 }),
  1: oscillator(0, {
    frequency: ['setValueAtTime', 660, currentTime + 1],
    stopTime: currentTime + 2,
  }),
})

And how to use it with linearRampToValueAtTime:

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: gain('output', { gain: 0.5 }),
  1: oscillator(0, {
    frequency: [
      ['setValueAtTime', 110, currentTime],
      ['linearRampToValueAtTime', 880, currentTime + 1],
    ],
    stopTime: currentTime + 2,
  }),
})

exponentialRampToValueAtTime:

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: gain('output', { gain: 0.5 }),
  1: oscillator(0, {
    frequency: [
      ['setValueAtTime', 110, currentTime],
      ['exponentialRampToValueAtTime', 880, currentTime + 1],
    ],
    stopTime: currentTime + 2,
  }),
})

setTargetAtTime:

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: gain('output', { gain: 0.5 }),
  1: oscillator(0, {
    frequency: [
      ['setValueAtTime', 110, currentTime],
      ['setTargetAtTime', 880, currentTime, 1],
    ],
    stopTime: currentTime + 2,
  }),
})

And finally setValueCurveAtTime:

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: gain('output', { gain: 0.5 }),
  1: oscillator(0, {
    frequency: [
      ['setValueCurveAtTime', Float32Array.of(440, 880, 110, 1760), currentTime, 2],
    ],
    stopTime: currentTime + 3,
  }),
})

Creating Custom Nodes

The audio graph can end up getting very large, repetitive and complicated so virtual-audio-graph gives a means of abstraction for creating encapsulated components that can be reused. These are called custom virtual audio nodes and are created with the createNode function like this:
const osc = createNode(({
  gain: gainValue,
  startTime,
  stopTime,
  ...rest,
}) => {
  const duration = stopTime - startTime
  return {
    0: gain('output', {
      gain: [
        ['setValueAtTime', 0, startTime],
        ['linearRampToValueAtTime', gainValue, startTime + duration * 0.15],
        ['setValueAtTime', gainValue, stopTime - duration * 0.25],
        ['linearRampToValueAtTime', 0, stopTime],
      ],
    }),
    1: oscillator(0, { startTime, stopTime, ...rest }),
  }
})

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: osc('output', {
    frequency: 110,
    gain: 0.2,
    startTime: currentTime,
    stopTime: currentTime + 1,
    type: 'square',
  }),
})

createNode takes a function that takes an object and returns a section of audio graph. This section of audio graph works in much the same way as the audio graph that is passed to virtualAudioGraph.update, but the special string "output" now outputs to whatever destinations the custom virtual audio node is connected to.

createNode returns a function that takes 2 arguments, just like the standard virtual audio node factory functions (e.g. oscillator and gain). The first argument represents the node output and the second is an object that is used to configure the section of audio graph as determined in the function passed to createNode.

Here is another example that builds upon the custom node we just created:

const oscBank = createNode(({
  frequency,
  ...rest,
}) => ({
  0: osc('output', {
    frequency,
    gain: 0.2,
    type: 'square',
    ...rest,
  }),
  1: osc('output', {
    detune: 7,
    frequency: frequency / 4,
    gain: 0.4,
    type: 'sawtooth',
    ...rest,
  }),
  2: osc('output', {
    gain: 0.1,
    detune: -4,
    frequency: frequency * 1.5,
    type: 'triangle',
    ...rest,
  }),
}))

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: oscBank('output', {
    frequency: 440,
    startTime: currentTime,
    stopTime: currentTime + 1,
  }),
})

In this way we can start to build up quite advanced graphs, but keep them organized and easy to understand.

Custom Nodes With Inputs

Sometimes you will want to connect a node to a custom node and will want to specify which nodes within the custom node that connection is made to. You can do this by passing the string "input" as the 3rd argument in the node constructor as below:

const pingPongDelay = createNode(({
  decay,
  delayTime,
}) => ({
  0: stereoPanner('output', { pan: -1 }),
  1: stereoPanner('output', { pan: 1 }),
  2: delay([1, 5], { delayTime, maxDelayTime: delayTime }),
  3: gain(2, { gain: decay }),
  4: delay([0, 3], { delayTime, maxDelayTime: delayTime }),
  5: gain(4, { gain: decay }, 'input'), // connections will be made here
}))

const { currentTime } = virtualAudioGraph

virtualAudioGraph.update({
  0: pingPongDelay('output', {
    decay: 0.8,
    delayTime: 0.25,
  }),
  1: oscillator([0, 'output'], { stopTime: currentTime + .2 }),
})

You can specify as many inputs as you like for custom virtual audio nodes and if an input node has no parameters you can pass null like this:

5: gain(4, null, 'input')

AudioWorklet Noise Generator

AudioWorklets are a fairly new addition to the Web Audio API so these demos won't work in all browsers!

If we have the following noise generator module at audioWorklets/noise.js:

class Noise extends AudioWorkletProcessor {
  static get parameterDescriptors () {
    return [{name: 'amplitude', defaultValue: 0.25, minValue: 0, maxValue: 1}]
  }

  process (inputs, [output], {amplitude}) {
    for (const outputChannel of output) {
      for (let i = 0; i < outputChannel.length; i++) {
        outputChannel[i] = 2 * (Math.random() - 0.5) * amplitude[i]
      }
    }

    return true
  }
}

registerProcessor('noise', Noise)

Then we can use it in virtual-audio-graph like this:

audioContext.audioWorklet.addModule('audioWorklets/noise.js')
  .then(() => {
    const noise = createWorkletNode('noise')

    virtualAudioGraph.update({
      0: noise('output', { amplitude: 0.25 }),
    })
  })

AudioWorklet Bit Crusher

Here is our bit crusher module at audioWorklets/bitCrusher.js:

class BitCrusher extends AudioWorkletProcessor {
  static get parameterDescriptors () {
    return [
      {name: 'bitDepth', defaultValue: 12, minValue: 1, maxValue: 16},
      {name: 'frequencyReduction', defaultValue: 0.5, minValue: 0, maxValue: 1},
    ]
  }

  constructor (options) {
    super(options)
    this.lastSampleValue = 0
    this.phase = 0
  }

  process ([input], [output], parameters) {
    const bitDepth = parameters.bitDepth
    const frequencyReduction = parameters.frequencyReduction
    for (let channel = 0; channel < input.length; channel++) {
      const inputChannel = input[channel]
      const outputChannel = output[channel]
      for (let i = 0; i < inputChannel.length; ++i) {
        const step = Math.pow(0.5, bitDepth[i])
        this.phase += frequencyReduction[i]
        if (this.phase >= 1) {
          this.phase -= 1
          this.lastSampleValue = step * Math.floor(inputChannel[i] / step + 0.5)
        }
        outputChannel[i] = this.lastSampleValue
      }
    }

    return true
  }
}

registerProcessor('bitCrusher', BitCrusher)

And here is how we can use it with virtual-audio-graph:

audioWorklet.addModule('audioWorklets/bitCrusher.js')
  .then(() => {
    const bitCrusher = createWorkletNode('bitCrusher')

    const { currentTime } = virtualAudioGraph

    virtualAudioGraph.update({
      0: bitCrusher('output', {
        bitDepth: 1,
        frequencyReduction: [
          ['setValueAtTime', 0.01, currentTime],
          ['linearRampToValueAtTime', 0.05, currentTime + 2],
          ['exponentialRampToValueAtTime', 0.01, currentTime + 4],
        ],
      }),
      1: oscillator(0, {
        frequency: 5000,
        stopTime: currentTime + 4,
        type: 'sawtooth',
      }),
    })
  })

Bringing It All Together

Here is a full working example that shows off a number of virtual-audio-graph's main features:

import createVirtualAudioGraph, {
  createNode,
  delay,
  gain,
  oscillator,
  stereoPanner,
} from 'virtual-audio-graph'

const osc = createNode(({
  gain: gainValue,
  startTime,
  stopTime,
  ...rest,
}) => {
  const duration = stopTime - startTime
  return {
    0: gain('output', {
      gain: [
        ['setValueAtTime', 0, startTime],
        ['linearRampToValueAtTime', gainValue, startTime + duration * 0.15],
        ['setValueAtTime', gainValue, stopTime - duration * 0.25],
        ['linearRampToValueAtTime', 0, stopTime],
      ],
    }),
    1: oscillator(0, { startTime, stopTime, ...rest }),
  }
})

const oscBank = createNode(({
  frequency,
  ...rest,
}) => ({
  0: osc('output', {
    frequency,
    gain: 0.2,
    type: 'square',
    ...rest,
  }),
  1: osc('output', {
    detune: 7,
    frequency: frequency / 4,
    gain: 0.4,
    type: 'sawtooth',
    ...rest,
  }),
  2: osc('output', {
    gain: 0.1,
    detune: -4,
    frequency: frequency * 1.5,
    type: 'triangle',
    ...rest,
  }),
}))

const pingPongDelay = createNode(({
  decay,
  delayTime,
}) => ({
  0: stereoPanner('output', { pan: -1 }),
  1: stereoPanner('output', { pan: 1 }),
  2: delay([1, 5], { delayTime, maxDelayTime: delayTime }),
  3: gain(2, { gain: decay }),
  4: delay([0, 3], { delayTime, maxDelayTime: delayTime }),
  5: gain(4, { gain: decay }, 'input'),
}))

const oscillators = createNode(({
  currentTime = virtualAudioGraph.currentTime,
  notes,
  noteLength,
}) => notes.reduce(
  (acc, frequency, i) => {
    const startTime = currentTime + noteLength * 2 * i
    acc[i] = oscBank('output', {
      frequency,
      startTime,
      stopTime: startTime + noteLength,
    })
    return acc
  },
  {}),
)

const chromaticScale = n => 440 * Math.pow(2, n / 12)
const noteLength = 0.075
const up = Array.apply(null, { length: 16 }).map((_, i) => chromaticScale(i))
const down = [...up].reverse()

virtualAudioGraph.update({
  1: pingPongDelay('output', {
    decay: 0.8,
    delayTime: noteLength * 1.55,
  }),
  2: gain(['output', 1], { gain: 0.25 }),
  3: oscillators(['output', 1], {
    noteLength,
    notes: [...up, ...down],
  }),
})

Thank you for reading and I hope you find this library useful. If you need any further help or have any feedback or suggestions you can get in touch via GitHub.

For the full documentation on the standard virtual audio node factories exported by virtual-audio-graph see here.

Happy coding!