cursoitaliano/tools/gen-audio.mjs

#!/usr/bin/env node
/**
 * Generate MP3 audio files from content/index.json using edge-tts.
 * Generates TWO variants per key:
 *   - <key>.mp3       (natural/fast speed, default rate)
 *   - <key>--slow.mp3 (slowed speech at -30% rate for learners)
 * Idempotent: skips files that already have >500 bytes.
 * Concurrency: max 4 parallel processes.
 * Retry: up to 2 retries per failure.
 * 
 * CLI: --only slow|fast|all (default: all)
 */

import { readFileSync, statSync, renameSync, unlinkSync, copyFileSync } from 'node:fs';
import { join, dirname } from 'node:path';
import { fileURLToPath } from 'node:url';
import { execFile } from 'node:child_process';
import { tmpdir } from 'node:os';
import { randomBytes } from 'node:crypto';

const __dirname = dirname(fileURLToPath(import.meta.url));
const ROOT = join(__dirname, '..');
const CONTENT_PATH = join(ROOT, 'content', 'index.json');
const AUDIO_DIR = join(ROOT, 'audio');

const VOICE_ELSA = 'it-IT-ElsaNeural';   // chunks + learner
const VOICE_DIEGO = 'it-IT-DiegoNeural'; // italian speaker

const MAX_CONCURRENCY = 4;
const MAX_RETRIES = 2;
const RETRY_DELAY_MS = 1000;

// Parse CLI flags
const args = process.argv.slice(2);
const onlyFlag = args.includes('--only') ? args[args.indexOf('--only') + 1] : 'all';
const MODE = onlyFlag === 'slow' ? 'slow' : onlyFlag === 'fast' ? 'fast' : 'all';

/**
 * Extract all unique {audioKey, text, voice} entries from index.json.
 */
function extractEntries() {
  const data = JSON.parse(readFileSync(CONTENT_PATH, 'utf8'));
  const seen = new Map(); // audioKey -> {text, voice}

  for (const mission of data.missions) {
    // Chunks: always Elsa
    if (mission.chunks) {
      for (const chunk of mission.chunks) {
        if (chunk.audioKey && !seen.has(chunk.audioKey)) {
          seen.set(chunk.audioKey, {
            text: chunk.it,
            voice: VOICE_ELSA,
          });
        }
      }
    }

    // Dialogues: voice depends on speaker
    if (mission.dialogues) {
      for (const dialogue of mission.dialogues) {
        if (!Array.isArray(dialogue)) continue;
        for (const line of dialogue) {
          if (line.audioKey && !seen.has(line.audioKey)) {
            const voice = line.speaker === 'italian' ? VOICE_DIEGO : VOICE_ELSA;
            seen.set(line.audioKey, {
              text: line.it,
              voice,
            });
          }
        }
      }
    }
  }

  return [...seen.entries()].map(([audioKey, { text, voice }]) => ({
    audioKey,
    text,
    voice,
  }));
}

/**
 * Check if a file exists and has size > minBytes.
 */
function hasAudio(filePath, minBytes = 0) {
  try {
    const st = statSync(filePath);
    return st.size > minBytes;
  } catch {
    return false;
  }
}

/**
 * Generate one audio file via edge-tts.
 * @param {string} outPath - Final output path
 * @param {string} text - Text to synthesize
 * @param {string} voice - Voice identifier
 * @param {string} [rate] - Optional rate adjustment (e.g., '-30%')
 * @returns {Promise<{ok: boolean, error?: string}>}
 */
async function generateAudio(outPath, text, voice, rate) {
  const tmpName = `tts_${randomBytes(6).toString('hex')}.mp3`;
  const tmpPath = join(tmpdir(), tmpName);

  const args = [
    '-m', 'edge_tts',
    '--voice', voice,
    '--text', text,
    '--write-media', tmpPath,
  ];

  if (rate) {
    // Use --rate=VALUE format to avoid argparse treating negative values as flags
    args.push(`--rate=${rate}`);
  }

  return new Promise((resolve) => {
    execFile('python', args, { timeout: 30000 }, (err, _stdout, stderr) => {
      if (err) {
        resolve({ ok: false, error: stderr || err.message });
        return;
      }

      // Verify tmp file was created with content
      try {
        const st = statSync(tmpPath);
        if (st.size < 500) {
          // Too small, likely failed silently
          try { unlinkSync(tmpPath); } catch {}
          resolve({ ok: false, error: `Output too small (${st.size} bytes)` });
          return;
        }
      } catch {
        resolve({ ok: false, error: 'Temp file not created' });
        return;
      }

      // Move to final location (copy+delete for cross-device safety)
      try {
        try {
          renameSync(tmpPath, outPath);
        } catch (renameErr) {
          if (renameErr.code === 'EXDEV') {
            copyFileSync(tmpPath, outPath);
            unlinkSync(tmpPath);
          } else {
            throw renameErr;
          }
        }
        resolve({ ok: true });
      } catch (moveErr) {
        try { unlinkSync(tmpPath); } catch {}
        resolve({ ok: false, error: `Move failed: ${moveErr.message}` });
      }
    });
  });
}

/**
 * Generate one audio variant (fast or slow) for a given key.
 * Returns { ok: true } or { ok: false, error: string }.
 */
async function generateOne(audioKey, text, voice, variant) {
  const suffix = variant === 'slow' ? '--slow' : '';
  const outPath = join(AUDIO_DIR, `${audioKey}${suffix}.mp3`);
  const rate = variant === 'slow' ? '-30%' : undefined;
  
  return generateAudio(outPath, text, voice, rate);
}

/**
 * Sleep helper.
 */
function sleep(ms) {
  return new Promise((r) => setTimeout(r, ms));
}

/**
 * Process entries with bounded concurrency.
 * Generates both fast and slow variants based on MODE flag.
 */
async function processAll(entries) {
  let generatedFast = 0;
  let generatedSlow = 0;
  let skippedFast = 0;
  let skippedSlow = 0;
  let failed = 0;
  const failures = [];

  // Build work items based on mode
  const todo = [];
  for (const entry of entries) {
    if (MODE === 'all' || MODE === 'fast') {
      const fastPath = join(AUDIO_DIR, `${entry.audioKey}.mp3`);
      if (hasAudio(fastPath, 500)) {
        skippedFast++;
      } else {
        todo.push({ ...entry, variant: 'fast' });
      }
    }
    if (MODE === 'all' || MODE === 'slow') {
      const slowPath = join(AUDIO_DIR, `${entry.audioKey}--slow.mp3`);
      if (hasAudio(slowPath, 500)) {
        skippedSlow++;
      } else {
        todo.push({ ...entry, variant: 'slow' });
      }
    }
  }

  console.log(`Mode: ${MODE}`);
  console.log(`Total keys: ${entries.length}`);
  if (MODE === 'all' || MODE === 'fast') {
    console.log(`Fast variant - Already exist (>500b): ${skippedFast}`);
  }
  if (MODE === 'all' || MODE === 'slow') {
    console.log(`Slow variant - Already exist (>500b): ${skippedSlow}`);
  }
  console.log(`To generate: ${todo.length}`);
  console.log(`Concurrency: ${MAX_CONCURRENCY}`);
  console.log('---');

  // Bounded concurrency pool
  let idx = 0;

  async function worker() {
    while (idx < todo.length) {
      const currentIdx = idx++;
      const item = todo[currentIdx];
      const suffix = item.variant === 'slow' ? '--slow' : '';
      const filePath = join(AUDIO_DIR, `${item.audioKey}${suffix}.mp3`);

      // Double-check (another worker may have done it)
      if (hasAudio(filePath, 500)) {
        if (item.variant === 'slow') skippedSlow++;
        else skippedFast++;
        continue;
      }

      let lastError = '';
      let success = false;

      for (let attempt = 0; attempt <= MAX_RETRIES; attempt++) {
        if (attempt > 0) {
          await sleep(RETRY_DELAY_MS * attempt);
        }
        const result = await generateOne(item.audioKey, item.text, item.voice, item.variant);
        if (result.ok) {
          success = true;
          break;
        }
        lastError = result.error;
      }

      if (success) {
        if (item.variant === 'slow') {
          generatedSlow++;
        } else {
          generatedFast++;
        }
        const total = generatedFast + generatedSlow;
        if (total % 20 === 0) {
          console.log(`  Progress: ${total}/${todo.length} generated (fast:${generatedFast}, slow:${generatedSlow})...`);
        }
      } else {
        failed++;
        failures.push({ key: `${item.audioKey}${suffix}`, error: lastError });
        console.log(`  FAIL: ${item.audioKey}${suffix} — ${lastError}`);
      }
    }
  }

  const workers = [];
  for (let w = 0; w < Math.min(MAX_CONCURRENCY, todo.length); w++) {
    workers.push(worker());
  }
  await Promise.all(workers);

  return { generatedFast, generatedSlow, skippedFast, skippedSlow, failed, failures };
}

// Main
async function main() {
  console.log('=== edge-tts Audio Generator (Dual Speed) ===');
  console.log(`Content: ${CONTENT_PATH}`);
  console.log(`Output:  ${AUDIO_DIR}`);
  console.log(`Mode:    ${MODE} (--only slow|fast|all)`);
  console.log('');

  const entries = extractEntries();
  console.log(`Unique audio keys extracted: ${entries.length}`);
  console.log('');

  const stats = await processAll(entries);

  console.log('');
  console.log('=== Summary ===');
  if (MODE === 'all' || MODE === 'fast') {
    console.log(`Fast (natural) - Generated: ${stats.generatedFast}, Skipped: ${stats.skippedFast}`);
  }
  if (MODE === 'all' || MODE === 'slow') {
    console.log(`Slow (-30%)    - Generated: ${stats.generatedSlow}, Skipped: ${stats.skippedSlow}`);
  }
  console.log(`Failed: ${stats.failed}`);

  if (stats.failures.length > 0) {
    console.log('');
    console.log('Failed keys:');
    for (const f of stats.failures) {
      console.log(`  ${f.key}: ${f.error}`);
    }
  }
}

main().catch((err) => {
  console.error('Fatal error:', err);
  process.exit(1);
});