AudioWorkletProcessor: process() method

inputs

An array of inputs connected to the node, each item of which is, in turn, an array of channels. Each channel is a Float32Array containing 128 samples. For example, inputs[n][m][i] will access n-th input, m-th channel of that input, and i-th sample of that channel.

Each sample value is in range of [-1 .. 1].

The number of inputs and thus the length of that array is fixed at the construction of the node (see AudioWorkletNode). If there is no active node connected to the n-th input of the node, inputs[n] will be an empty array (zero input channels available).

The number of channels in each input may vary, depending on channelCount and channelCountMode properties.

outputs

An array of outputs that is similar to the inputs parameter in structure. It is intended to be filled during the execution of the process() method. Each of the output channels is filled with zeros by default — the processor will output silence unless the output arrays are modified.

parameters

An object containing string keys and Float32Array values. For each custom AudioParam defined using the parameterDescriptors getter, the key in the object is a name of that AudioParam, and the value is a Float32Array. The values of the array are calculated by taking scheduled automation events into consideration.

If the automation rate of the parameter is "a-rate", the array will contain 128 values — one for each frame in the current audio block. If there's no automation happening during the time represented by the current block, the array may contain a single value that is constant for the entire block, instead of 128 identical values.

If the automation rate is "k-rate", the array will contain a single value, which is to be used for each of 128 frames.

A Boolean value indicating whether or not to force the AudioWorkletNode to remain active even if the user agent's internal logic would otherwise decide that it's safe to shut down the node.

The returned value lets your processor have influence over the lifetime policy of the AudioWorkletProcessor and the node that owns it. If the combination of the return value and the state of the node causes the browser to decide to stop the node, process() will not be called again.

Returning true forces the Web Audio API to keep the node alive, while returning false allows the browser to terminate the node if it is neither generating new audio data nor receiving data through its inputs that it is processing.

The 3 most common types of audio node are:

1. A source of output. An AudioWorkletProcessor implementing such a node should return true from the process method as long as it produces an output. The method should return false as soon as it's known that it will no longer produce an output. For example, take the AudioBufferSourceNode — the processor behind such a node should return true from the process method while the buffer is playing, and start returning false when the buffer playing has ended (there's no way to call play on the same AudioBufferSourceNode again).
2. A node that transforms its input. A processor implementing such a node should return false from the process method to allow the presence of active input nodes and references to the node to determine whether it can be garbage-collected. An example of a node with this behavior is the GainNode. As soon as there are no inputs connected and references retained, gain can no longer be applied to anything, so it can be safely garbage-collected.
3. A node that transforms its input, but has a so-called tail-time — this means that it will produce an output for some time even after its inputs are disconnected or are inactive (producing zero-channels). A processor implementing such a node should return true from the process method for the period of the tail-time, beginning as soon as inputs are found that contain zero-channels. An example of such a node is the DelayNode — it has a tail-time equal to its delayTime property.

As the process() method is implemented by the user, it can throw anything. If an uncaught error is thrown, the node will emit an processorerror event and will output silence for the rest of its lifetime.

In this example we create an AudioWorkletProcessor that outputs white noise to its first output. The gain can be controlled by the customGain parameter.

class WhiteNoiseProcessor extends AudioWorkletProcessor {
  process(inputs, outputs, parameters) {
    // take the first output
    const output = outputs[0];
    // fill each channel with random values multiplied by gain
    output.forEach((channel) => {
      for (let i = 0; i < channel.length; i++) {
        // generate random value for each sample
        // Math.random range is [0; 1); we need [-1; 1]
        // this won't include exact 1 but is fine for now for simplicity
        channel[i] =
          (Math.random() * 2 - 1) *
          // the array can contain 1 or 128 values
          // depending on if the automation is present
          // and if the automation rate is k-rate or a-rate
          (parameters["customGain"].length > 1
            ? parameters["customGain"][i]
            : parameters["customGain"][0]);
      }
    });
    // as this is a source node which generates its own output,
    // we return true so it won't accidentally get garbage-collected
    // if we don't have any references to it in the main thread
    return true;
  }
  // define the customGain parameter used in process method
  static get parameterDescriptors() {
    return [
      {
        name: "customGain",
        defaultValue: 1,
        minValue: 0,
        maxValue: 1,
        automationRate: "a-rate",
      },
    ];
  }
}

This is not a method provided by browsers, but a callback method that must be written in client code.

• Using the Web Audio API