MuLawToWavFileEncoderStream.ts

import { TransformStream } from "stream/web"

// Based on https://github.com/jremi/twilio-media-stream-save-audio-file/blob/69141d4ca4de3807dae5282ce5e600b110592f40/index.js
// and on https://en.wikipedia.org/wiki/WAV#WAV_file_header

const textEncoder = new TextEncoder()
const text = (input: string) => textEncoder.encode(input)
const int16 = (value: number) => new Uint8Array([value & 0xff, (value >> 8) & 0xff])
const int32 = (value: number) =>
  new Uint8Array([
    value & 0xff,
    (value >> 8) & 0xff,
    (value >> 16) & 0xff,
    (value >> 24) & 0xff,
  ])

interface WavHeaderInput {
  numberOfChannels: number
  sampleRateHz: number
  bitsPerSample: number
  dataLength: number
}
const DEFAULT_WAV_HEADER_INPUT: WavHeaderInput = {
  numberOfChannels: 1,
  sampleRateHz: 8_000,
  bitsPerSample: 8,
  dataLength: Infinity,
}
function saturateInt32(value: number) {
  return value < 0 ? 0 : value > 0xffffffff ? 0xffffffff : value
}
function getWavHeader(options: WavHeaderInput, headerByteLength?: number): Uint8Array {
  const resolvedHeaderByteLength = headerByteLength ?? getWavHeader(options, 0).length
  return new Uint8Array([
    // [Master RIFF chunk]

    // 1 - 4 = "RIFF"
    ...text("RIFF"),
    // 5 - 8 = File size - Overall file size minus 8 bytes
    ...int32(saturateInt32(options.dataLength + resolvedHeaderByteLength - 8)),
    // 9 - 12 = "WAVE"
    ...text("WAVE"),

    // [Chunk describing the data format]

    // 13 - 16 = "fmt "
    ...text("fmt "),
    // 17 - 20 = Chunk size minus 8 bytes, which is 16 bytes here  (0x10)
    ...int32(16),
    // 21 - 22 = Audio format (1: PCM integer, 3: IEEE 754 float()
    ...int16(7),
    // 23 - 24 = Number of channels
    ...int16(options.numberOfChannels),
    // 25 - 28 = Sample rate (in hertz)
    ...int32(options.sampleRateHz),
    // 29 - 32 = Number of bytes to read per second = (Sample Rate * BitsPerSample * Channels) / 8 = (Frequency * BytePerBloc)
    ...int32(
      (options.sampleRateHz * options.bitsPerSample * options.numberOfChannels) / 8
    ),
    // 33 - 34 = BytePerBloc = (NbrChannels * BitsPerSample / 8).
    ...int16((options.numberOfChannels * options.bitsPerSample) / 8),
    // 35 - 36 = Bits per sample
    ...int16(options.bitsPerSample),

    // [Data chunk]

    // 37 - 40 = "data"
    ...text("data"),
    // 41 - 44 = Size of the data section.
    ...int32(saturateInt32(options.dataLength)),
  ])
}

export default class WavFileEncoderStream extends TransformStream<
  Uint8Array,
  Uint8Array
> {
  constructor(options: Partial<WavHeaderInput> = {}) {
    super({
      start(controller) {
        controller.enqueue(
          getWavHeader({
            ...DEFAULT_WAV_HEADER_INPUT,
            ...options,
          })
        )
      },
      transform(chunk, controller) {
        controller.enqueue(chunk)
      },
    })
  }
}