web-audio-beat-detector

A beat detection utility which is using the Web Audio API.

This module is based on the technique explained by Joe Sullivan in his article Beat Detection Using JavaScript and the Web Audio API. It retrieves the beats as BPM of a given AudioBuffer. The algorithm used is not as complex (and expensive to compute) as many others. But it yields surprisingly good results especially for electronic music.

Usage

The web-audio-beat-detector module is available on npm and can be installed as usual.

npm install web-audio-beat-detector

You can then import its public function analyze() like this:

import { analyze } from 'web-audio-beat-detector';

The analyze() function expects an AudioBuffer as its first parameter and it returns a Promise which eventually resolves with the tempo of that buffer as a number. An example usage might look like this:

analyze(audioBuffer)
    .then((tempo) => {
        // the tempo could be analyzed
    })
    .catch((err) => {
        // something went wrong
    });

Additionally you can also import the guess() function like this:

import { guess } from 'web-audio-beat-detector';

The guess() function expects an AudioBuffer as well and also returns a Promise. The Promise will resolve with an object containing the estimated BPM (the rounded tempo) and the offset of the first beat in seconds.

guess(audioBuffer)
    .then(({ bpm, offset, tempo }) => {
        // the bpm and offset could be guessed
        // the tempo is the same as the one returned by analyze()
    })
    .catch((err) => {
        // something went wrong
    });

analyze() and guess() do both support offset and duration as optional arguments. When specified these two values are used to select only a part of the given AudioBuffer. There usage is the same as described in the documentation of the AudioBufferSourceNode.start() method.

By default the bpm are expected to be between 90 and 180 bpm. This can be changed by specifying the tempoSettings. These can be passed as the last argument to analyze() or guess().

// with an offset and duration
analyze(audioBuffer, 1, 10, { maxTempo: 120, minTempo: 60 });
// with no other arguments
analyze(audioBuffer, { maxTempo: 120, minTempo: 60 });

Acknowledgement

A more comprehensive implementation has been done by José M. Pérez. It comes with an UI to search for tracks on Spotify which can then be analyzed. He also wrote a blog post (Detecting tempo of a song using browser's Audio API) about it.