My TTS read a whole novel. The model was the most reliable part of the pipeline
Описание и применение
В проекте TTS для кримано‑татарского языка ошибки возникали из‑за обработки текста и аудио, а не модели. Устранение объединения коротких предложений, улучшение обрезки дыхания и корректировка пауз снизили процент ошибок до 1,84 %.
Задача
The TTS pipeline produced errors due to text and audio preprocessing steps rather than the model itself.
Результат
After correcting preprocessing bugs, the TTS pipeline produced a 1 hour 58 minute audio of a 16‑chapter Crimean Tatar novel with a character error rate of 1.84%, demonstrating that model quality was high and issues lay in text handling.
Рабочий процесс
- 1. Verify text is not in training data (zero 6‑word matches).
- 2. Render full chapter without gluing short sentences.
- 3. Benchmark post‑processing steps (gluing, breath trimming, pause insertion) against doing nothing.
- 4. Replace breath trimmer with loudness‑based trimming.
- 5. Adjust pause durations based on human narrator statistics.
- 6. Re‑render chapters and compare character error rates.
- 7. Run extraction and guard checks on full book to catch formatting bugs.
Доказательства
The model held up. The problems weren't in the model but in the text and audio preparation around it
On glued chunks the model started throwing out entire sentences
The chapter scored better with no trimming at all than with my "fix"
- Дата записи: 2026-10-09T18:55:51.397Z