Download nircmdc, ffmpeg and deno, run using "deno run -A microphone-keyboard.js"

microphone-keyboard.js

import FFT from 'https://raw.githubusercontent.com/qgustavor/stream-audio-fingerprint/master/src/lib/fft.ts'

let device = Deno.args[0]
if (!device) {
  const devicesProcess = await Deno.run({
    cmd: ['ffmpeg', '-list_devices', 'true', '-f', 'dshow', '-i', 'dummy'],
    stderr: 'piped'
  })
  const devicesOutput = new TextDecoder().decode(await devicesProcess.stderrOutput())
  const devices = Array.from(devicesOutput.matchAll(/\[dshow.* "(.*)" \(audio\)/g)).map(e => e[1])

  devices.sort((a, b) => {
    if (a.match(/microphone/i)) return -1
    if (b.match(/microphone/i)) return 1
    return a.localeCompare(b)
  })

  device = devices[0]
}

console.log('Using device', device)
let handler

async function startListening () {
  const listenerProcess = await Deno.run({
    cmd: ['ffmpeg', '-f', 'dshow', '-i', 'audio=' + device, '-ar', '22050', '-ac', '1', '-f', 's16le', '-acodec', 'pcm_s16le', '-'],
    stdout: 'piped',
    stderr: 'null'
  })

  const nfft = 512
  const step = nfft / 2
  const bps = 2
  const fft = new FFT(nfft)
  const hwin = Array(nfft).fill(null).map((_f, i) => (
    0.5 * (1 - Math.cos(((2 * Math.PI) * i) / (nfft - 1)))
  ))
  let buffer = new Uint8Array(0)
  let bufferDelta = 0
  let stepIndex = 0
  let lastMatchedIndex
  let lastMatchedTime = Date.now()
  let timeBetweenMatches = 500

  let minVolume = Number(Deno.args[1]) || 0.01

  for await (const chunk of listenerProcess.stdout.readable) {
    const concatedBuffer = new Uint8Array(buffer.length + chunk.length)
    concatedBuffer.set(buffer, 0)
    concatedBuffer.set(chunk, buffer.length)
    buffer = concatedBuffer

    const bufferView = new DataView(concatedBuffer.buffer)

    while ((stepIndex + nfft) * bps < buffer.length + bufferDelta) {
      const data = new Array(nfft) // window data
      const image = new Array(nfft).fill(0)

      // fill the data, windowed (HWIN) and scaled
      for (let i = 0, limit = nfft; i < limit; i++) {
        const readInt = bufferView.getInt16((stepIndex + i) * bps - bufferDelta, true)
        data[i] = (hwin[i] * readInt) / Math.pow(2, 8 * bps - 1)
      }
      stepIndex += step

      fft.forward(data, image) // compute FFT

      const largestValue = fft.spectrum.reduce((a, b) => a > b ? a : b)
      if (largestValue > minVolume) {
        const largestIndex = fft.spectrum.indexOf(largestValue)
        if (lastMatchedIndex !== largestIndex || lastMatchedTime + timeBetweenMatches < Date.now()) {
          lastMatchedTime = Date.now()
          lastMatchedIndex = largestIndex
          if (handler) handler(largestIndex)
        }
      }
    }
  }
}

startListening()

console.log('--- Configuring frequencies ---')
const keys = {}
while (true) {
  const key = prompt('Press some key to be emulated or press enter to continue:')
  if (!key) break
  console.log('Make a sound:')
  const value = await new Promise(resolve => { handler = resolve })
  console.log('Frequency %s registered', value)
  keys[value] = key
}

console.log('--- Emulating keys ---')
while (true) {
  const value = await new Promise(resolve => { handler = resolve })
  const key = keys[value]
  console.log('Got frequency %s key %s', value, key)
  if (!key) continue

  await Deno.run({
    cmd: ['nircmdc.exe', 'sendkey', key, 'press'],
  })
}