Web Audio API example

Web Audio API visualizer. AnalyserNode FFT from the microphone, drawn on canvas. Frequency data the audio element will not give you.

The problem

An <audio> element plays sound. It will not tell you what the signal is doing. Frequency and amplitude stay inside the player.

The method

AudioContext is a graph. A source feeds a GainNode, then an AnalyserNode. The analyser runs an FFT. getByteFrequencyData() is the spectrum. getByteTimeDomainData() is the waveform. Same node, two reads.

What to do

  1. Run AudioContext. That is the graph.
  2. Connect the mic, or pick a file and connect that.
  3. Run the analyser. Then switch to the time domain.

AudioContext

Creates the graph. Browsers keep it suspended until a user gesture. resume() is that yes.

const ctx = new AudioContext()
await ctx.resume()

createMediaStreamSource()

Puts the microphone on the graph. Do not connect a mic to destination or you get feedback.

const stream = await navigator.mediaDevices.getUserMedia({ audio: true })
const source = ctx.createMediaStreamSource(stream)
source.connect(gain)

createMediaElementSource()

Puts an <audio> element on the graph. One element can only make one source node. Connect through to destination or you will not hear it.

const source = ctx.createMediaElementSource(audio)
source.connect(gain)
gain.connect(ctx.destination)

AnalyserNode

An FFT on the graph. fftSize is the window. frequencyBinCount is half of that. getByteFrequencyData() fills one byte per bin. Gain and fftSize are the node, not a CSS scale.

const analyser = ctx.createAnalyser()
analyser.fftSize = 256
gain.connect(analyser)
analyser.getByteFrequencyData(bins)
0.8

getByteTimeDomainData()

Same analyser, other buffer. 128 is silence. The line is the waveform, not the spectrum.

analyser.getByteTimeDomainData(bins)

All Web API labs

  • HTML usermedia element example
  • MediaRecorder API example
  • Web Speech API example