rhinelab-ui/scripts/extract-typing-audio.mjs

// Reproduce typing clicks from the user's local reference video.
// node scripts/extract-typing-audio.mjs path/to/ffmpeg.exe
// These short excerpts retain the source recording's provenance, not the score's license.
import fs from "node:fs";
import crypto from "node:crypto";
import { execFileSync } from "node:child_process";
const ffmpeg = process.argv[2],
  video = fs.readdirSync(".").find((x) => x.endsWith(".mp4"));
if (!ffmpeg || !video)
  throw Error("A local reference MP4 and FFmpeg path are required");
const rate = 48000,
  start = 6.84,
  seconds = 0.55;
const bytes = execFileSync(
  ffmpeg,
  [
    "-v",
    "error",
    "-ss",
    String(start),
    "-i",
    video,
    "-t",
    String(seconds),
    "-vn",
    "-af",
    "pan=mono|c0=0.5*c0+0.5*c1",
    "-ar",
    String(rate),
    "-f",
    "f32le",
    "pipe:1",
  ],
  { maxBuffer: 1024 * 1024 },
);
const source = new Float32Array(
  bytes.buffer,
  bytes.byteOffset,
  bytes.length / 4,
);
const cuts = [6.864, 6.906, 6.948],
  duration = 0.038;
const samples = cuts.map((time) =>
  source.slice(
    Math.round((time - start) * rate),
    Math.round((time - start + duration) * rate),
  ),
);
let peak = 0;
for (const sample of samples) {
  const dc = sample.reduce((sum, x) => sum + x, 0) / sample.length;
  for (let i = 0; i < sample.length; i++) {
    const fadeIn = Math.min(1, i / (rate * 0.00035)),
      fadeOut = Math.min(1, (sample.length - 1 - i) / (rate * 0.0035));
    sample[i] = (sample[i] - dc) * fadeIn * fadeOut;
    peak = Math.max(peak, Math.abs(sample[i]));
  }
}
// Shared gain preserves the three original clicks' relative accents.
const gain = 0.3 / peak;
const pcm = samples.map((sample) => {
  const out = Buffer.alloc(sample.length * 2);
  for (let i = 0; i < sample.length; i++)
    out.writeInt16LE(Math.round(sample[i] * gain * 32767), i * 2);
  return out;
});
fs.mkdirSync("public/audio", { recursive: true });
function wav(data, file) {
  const out = Buffer.alloc(44 + data.length);
  out.write("RIFF");
  out.writeUInt32LE(out.length - 8, 4);
  out.write("WAVEfmt ", 8);
  out.writeUInt32LE(16, 16);
  out.writeUInt16LE(1, 20);
  out.writeUInt16LE(1, 22);
  out.writeUInt32LE(rate, 24);
  out.writeUInt32LE(rate * 2, 28);
  out.writeUInt16LE(2, 32);
  out.writeUInt16LE(16, 34);
  out.write("data", 36);
  out.writeUInt32LE(data.length, 40);
  data.copy(out, 44);
  fs.writeFileSync(file, out);
}
// Review-only source excerpt. Uniform gain makes its level comparable to the rebuilt sequence.
const referencePcm = Buffer.alloc(source.length * 2);
for (let i = 0; i < source.length; i++)
  referencePcm.writeInt16LE(
    Math.round(Math.max(-1, Math.min(1, source[i] * gain * 0.2)) * 32767),
    i * 2,
  );
wav(referencePcm, "reference/typing-original.wav");
const preview = Buffer.alloc(rate * 2 * 2);
const pulseTimes = [
  0.02, 0.064, 0.108, 0.152, 0.2, 0.244, 0.288, 0.332, 0.38, 0.424, 0.468,
  0.516, 1.02, 1.068, 1.116, 1.16, 1.208, 1.256, 1.304, 1.352,
];
pulseTimes.forEach((t, n) => {
  const click = pcm[n % pcm.length],
    offset = Math.round(t * rate);
  for (let i = 0; i < click.length / 2; i++)
    preview.writeInt16LE(
      Math.round(click.readInt16LE(i * 2) * 0.2),
      (offset + i) * 2,
    );
});
wav(preview, "public/audio/typing-preview.wav");
fs.writeFileSync(
  "src/typing-samples.ts",
  `// Generated by scripts/extract-typing-audio.mjs. Source provenance: public/audio/typing-source.json.\nexport const TYPING_SAMPLE_RATE = ${rate};\nexport const TYPING_PCM = ${JSON.stringify(
    pcm.map((x) => x.toString("base64")),
    null,
    2,
  )} as const;\n`,
);
const report = {
  source: video,
  sourceSha256: crypto
    .createHash("sha256")
    .update(fs.readFileSync(video))
    .digest("hex"),
  sampleRate: rate,
  channels: 1,
  cuts: cuts.map((start) => ({ start, end: +(start + duration).toFixed(3) })),
  processing: {
    dcRemoval: true,
    fadeInMs: 0.35,
    fadeOutMs: 3.5,
    sharedGain: gain,
    denoising: false,
    pitchChange: false,
  },
  scope:
    "Three short excerpts from user-provided reference. Not original synthesis; source rights remain with the original creators.",
};
fs.writeFileSync(
  "public/audio/typing-source.json",
  JSON.stringify(report, null, 2) + "\n",
);
console.log(
  JSON.stringify(
    { cuts: report.cuts, bytes: pcm.reduce((n, b) => n + b.length, 0), gain },
    null,
    2,
  ),
);