Made of little things.

// Prints how Piper/espeak-ng will pronounce every radio clip (IPA phonemes per sentence),
// for checking pronunciation. Run with: bun tools/voices/phonemes.ts
import { homedir } from 'node:os';
import { join } from 'node:path';
import { CLIPS } from '../../src/game/radio-script.ts';

const dir = join(homedir(), '.cache/piper-dl');
const keys = Object.keys(CLIPS).filter((k) => !/\.(speed|fast)\d+$/.test(k) || /(speed|fast)(60|100|110|140|160)$/.test(k));
const input = keys.map((k) => JSON.stringify({ text: CLIPS[k].say, output_file: join(dir, 'phon.wav') })).join('\n') + '\n';
const p = Bun.spawnSync([join(dir, 'piper/piper/piper.exe'), '--model', join(dir, 'models/en_US-libritts-high.onnx'), '--json-input', '--debug'],
  { stdin: Buffer.from(input), cwd: join(dir, 'piper/piper') });
const log = p.stderr.toString() + p.stdout.toString();
// Piper logs "Phonemizing text: ..." per input, then "Converting N phoneme(s) to ids: ..." per sentence.
const out: string[] = [];
for (const line of log.split(/\r?\n/)) {
  const text = line.match(/Phonemizing text: (.*)$/);
  if (text) out.push(`\n${keys[out.filter((l) => l.startsWith('\n')).length] ?? '?'}: ${text[1].trim()}`);
  const ph = line.match(/to ids: (.*)$/);
  if (ph) out.push(`    ${ph[1].trim()}`);
}
console.log(out.join('\n'));