audio-bench

Slice any browser-decodable audio into ASR-friendly mono 16 kHz WAV chunks. Zero dependencies · pure WebAudio · works offline.

Part of audio-bench (drag-and-drop web tool). This package is the slicing core as a library.

Install

npm install audio-bench

Use

import { processAudio } from "audio-bench";

// file: any Blob the browser can decode (mp3 / m4a / wav / flac / ogg…)
const { durationSec, chunks } = await processAudio(file, {
  chunkSec: 85,            // default 85 s — safe for virtually every ASR
  sampleRate: 16000,       // default 16 kHz mono
  onProgress: (p) => console.log(`${p.phase} ${p.done}/${p.total}`),
});

for (const c of chunks) {
  // c.blob is a 16-bit PCM WAV (mono, ≤ chunkSec long) — ready for any ASR
  await asr(c.blob);
}

Pure helpers also exported (Node-safe, no WebAudio needed):

import { encodeWavBytes, sliceSamples } from "audio-bench";

const chunks = sliceSamples(monoFloat32, 16000, 85); // → [{index,startSec,endSec,samples}]
const wav = encodeWavBytes(chunks[0].samples, 16000); // → Uint8Array

processAudio uses AudioContext / OfflineAudioContext, so it runs in browsers (and Electron). For Node pipelines, decode with your own decoder and use the pure helpers above.

API

| Export | Kind | Notes | |--------|------|-------| | processAudio(file, opts?) | browser | Blob → {durationSec, chunkSec, sampleRate, channels, chunks[]} | | sliceSamples(samples, rate, chunkSec) | pure | fixed-length sample slicing | | encodeWavBytes(samples, rate) | pure | mono Float32 → 16-bit PCM WAV Uint8Array | | CHUNK_SEC | const | 85 |

License

MIT