Use essentia bpm

Author: sliterok

package.json

@@ -53,7 +53,6 @@
 		"dotenv": "^16.4.5",
 		"express": "^4.18.2",
                 "fft-js": "^0.0.12",
-                "music-tempo": "^1.0.3",
                 "mic": "^2.1.1",
                 "essentia.js": "^0.1.3",
 		"react": "^19.1.0",

pnpm-lock.yaml

@@ -1,6 +1,6 @@
 import { WebSocketServer } from 'ws'
 import fftjs from 'fft-js'
-import MusicTempo from 'music-tempo'
+import { Essentia, EssentiaWASM } from 'essentia.js'
 // eslint-disable-next-line import/default
 import RingBufferTs from 'ring-buffer-ts'
 const RingBuffer = RingBufferTs.RingBuffer
@@ -29,8 +29,9 @@ const sampleRateDefault = 44100
 const bpmWindow = sampleRateDefault * 8
 const sampleBuffer = new RingBuffer<number>(bpmWindow)
 let lastBpmUpdate = 0
+const essentiaPromise = Promise.resolve(new Essentia(EssentiaWASM))
 
-export function processAudio(buffer: Buffer, sampleRate = sampleRateDefault) {
+export async function processAudio(buffer: Buffer, sampleRate = sampleRateDefault) {
 	const samples = new Int16Array(buffer.buffer, buffer.byteOffset, buffer.byteLength / 2)
 	const input = Array.from(samples, s => s / 32768)
 	for (const s of input) sampleBuffer.add(s)
@@ -56,8 +57,10 @@ export function processAudio(buffer: Buffer, sampleRate = sampleRateDefault) {
 	if (now - lastBpmUpdate > 2000 && sampleBuffer.getBufferLength() >= sampleRate * 4) {
 		lastBpmUpdate = now
 		try {
-			const mt = new MusicTempo(Float32Array.from(sampleBuffer.toArray()))
-			const tempo = parseFloat(String(mt.tempo))
+			const essentia = await essentiaPromise
+			const signal = essentia.arrayToVector(Float32Array.from(sampleBuffer.toArray()))
+			const res = essentia.RhythmExtractor2013(signal, 208, 'multifeature', 40)
+			const tempo = res.bpm
 			if (!Number.isNaN(tempo)) audioState.bpm = tempo
 		} catch {
 			// ignore errors

src/backend/wsAudio.ts

@@ -0,0 +1 @@
+declare module 'essentia.js'

src/types/essentia.js.d.ts

@@ -24,16 +24,16 @@ describe('processAudio', () => {
         beforeEach(() => {
                 resetAudioState()
         })
-       test('detects frequency', () => {
+       test('detects frequency', async () => {
                const buf = genSine(440, 1234, 44100)
-               processAudio(buf)
+               await processAudio(buf)
                expect(audioState.freq).toBeGreaterThan(430)
                expect(audioState.freq).toBeLessThan(450)
        })
 
-       test('detects bpm', () => {
+       test('detects bpm', async () => {
                const buf = genBeat(120, 4, 44100)
-               processAudio(buf)
+               await processAudio(buf)
                expect(audioState.bpm).toBeGreaterThan(110)
                expect(audioState.bpm).toBeLessThan(130)
        })