Recording

1// recording.mjs — the person's own voice recording as a voice line: converted to the engine's wav,
2// time-stretched by `speed` with the pitch kept, with word timings scaled to match.
3import { spawnSync } from "node:child_process";
4import { existsSync, mkdirSync } from "node:fs";
5import { dirname } from "node:path";
6
7// ffmpeg's atempo takes 0.5..2 per stage; chain stages for the rest of the 0.1..3 speed range.
8export function atempoChain(speed) {
9  const stages = [];
10  let left = speed;
11  while (left > 2) {
12    stages.push(2);
13    left /= 2;
14  }
15  while (left < 0.5) {
16    stages.push(0.5);
17    left /= 0.5;
18  }
19  if (Math.abs(left - 1) > 1e-9 || stages.length === 0) stages.push(left);
20  return stages.map((s) => `atempo=${Number(s.toFixed(6))}`).join(",");
21}
22
23/** src -> 44.1k mono wav at wavAbs, played `speed` times faster. Returns { ok, error? }. */
24export function stretchToWav(src, wavAbs, speed = 1) {
25  if (!existsSync(src)) return { ok: false, error: `recording not found: ${src}` };
26  mkdirSync(dirname(wavAbs), { recursive: true });
27  const filter = speed === 1 ? [] : ["-af", atempoChain(speed)];
28  const r = spawnSync(
29    "ffmpeg",
30    ["-y", "-loglevel", "error", "-i", src, ...filter, "-ar", "44100", "-ac", "1", wavAbs],
31    { encoding: "utf8" },
32  );
33  if (r.error || r.status !== 0 || !existsSync(wavAbs)) {
34    return { ok: false, error: `ffmpeg could not convert ${src}: ${r.error?.message ?? r.stderr}` };
35  }
36  return { ok: true };
37}
38
39/** Caller-supplied word timings ({text|word, start, end}) as [{text,start,end}], or null when absent. */
40export function parseWords(words) {
41  if (!Array.isArray(words) || words.length === 0) return null;
42  return words.map((w, i) => {
43    const text = w?.text ?? w?.word;
44    if (typeof text !== "string" || !Number.isFinite(w.start) || !Number.isFinite(w.end)) {
45      throw new Error(`words[${i}] needs text (or word), start and end in seconds`);
46    }
47    return { text, start: w.start, end: w.end };
48  });
49}
50
51/** Words of a recording played `speed` times faster land at start/speed. */
52export const scaleWords = (words, speed) =>
53  words.map((w) => ({ ...w, start: w.start / speed, end: w.end / speed }));