opus.js

Inspired by OpusScript and opus-recorder,a JavaScript library compiled from libopus using Emscripten.

Usage

Install with

npm i -S @vocaltale/opus.js

// or

yarn add @vocaltale/opus.js

Here is an example using Node.js to encode PCM in 32-bits floating point to opus and write it to a file using Ogg container format.

const fs = require('fs')
const {
  Encoder,
  Ogg,
} = require('@vocaltale/opus.js')

const encoder = new Encoder(48000, 1, { maxFrameSize: 960 })
// const decoder = new Decoder(48000, 1, { maxFrameSize: 960 })

const raw = new Float32Array(960)
let encoded = null
const ogg = new Ogg(48000, 1, 960, 16)
const stream = fs.createWriteStream('test.opus')

stream.write(ogg.generateIdPage())
stream.write(ogg.generateCommentPage())

raw.fill(0.1)
encoded = encoder.encodeFloat(raw)

// console.log(encoded.length)
const packets = ogg.segmentPacket(encoded)

packets.forEach(packet => {
  stream.write(packet)
})

const all = ogg.finalize()

stream.write(all)
stream.end()

encoder.delete()

For browser applications, we might want to use esModule version for Webpack 5 Lazy-loading.

import {
  Codec,
  OpusApplication
} from '@vocaltale/opus.js/esm'

const codec = new Codec(48000, 1, {
  maxFrameSize: 960,
  application: OpusApplication.RESTRICTED_LOWDELAY,
})

const raw = new Float32Array(960)
let encoded = null

raw.fill(0.1)
encoded = codec.encodeFloat(raw)

// ...

const decoded = codec.decodeFloat(encoded)

codec.delete()

The only tricky part is to place the build/opus-<codec|encoder|decoder>.wasm to the correct place so the script can run, but let's stop here for you as a challenge.

Remember opus is a stateful codec; don't use the same encoder/decoder with multiple sources. Or you might can weird, noisy results.

Enhancements

1. Webpack 5 Lazy-loading support (using await import() syntax) (only available for esModule, @vocaltale/opus.js/esm)
2. Automatically fallback to opus-<codec|encoder|decoder>.asm.js if WASM is not supported
3. Support 32-bits Float formatted PCM data
4. Lower memory usage by limiting the stack size and the heap size further
5. Separated encoder and decoder, .wasm files can be much smaller if don't need both of them. Original codec is also there if you need both of them; they share the same stack and heap memory.