Word timings

final result = await tts.generate('Highlight this sentence as it plays.');

for (final timing in result.wordTimings) {
  print('${timing.word}: ${timing.startTime}–${timing.endTime}');
}

Start the UI timer when playback actually begins:

await tts.play(
  result,
  AudioPlayOptions(
    onPlaybackStart: () {
      final startedAt = DateTime.now();
      timer = Timer.periodic(const Duration(milliseconds: 50), (_) {
        final seconds =
            DateTime.now().difference(startedAt).inMilliseconds / 1000;
        final active = result.wordTimings
            .where((word) => seconds >= word.startTime && seconds < word.endTime)
            .firstOrNull;
        setActiveWordIndex(active?.wordIndex);
      });
    },
  ),
);

Timings are inferred from predicted phoneme duration output. They are designed for read-aloud UI, not forensic forced alignment.

• Prefer sentences and short paragraphs.
• Treat indexes as local to the current result.
• Timings can be empty when duration output is unavailable.
• Use stream() for long input and switch highlighting data per chunk.