Making open-source text to speech sound right: ranked voices, text normalization and one engine for Python and the browser

Описание и применение

loudness-tts — открытая система синтеза речи, которая автоматически выбирает лучшую лицензированную голосовую модель, применяет правила нормализации текста для английского и испанского, нормализует громкость до –16 LUFS и генерирует точные субтитры, доступна как для Python, так и для JavaScript.

Задача

Open-source TTS models run fast but often sound unnatural due to voice quality, lack of text normalization, and unoptimized audio output.

Результат

loudness-tts delivers natural-sounding speech with accurate text normalization, licensed voice selection, and publish‑ready audio and subtitles across Python and JavaScript platforms.

Рабочий процесс

• 1. Install loudness-tts via pip or use the browser demo.
• 2. Run the TTS command with text input and optional output options (e.g., –o speech.wav --srt speech.srt).
• 3. For chat applications, launch the OpenAI‑compatible server with loudness-tts serve or use the Docker image.
• 4. Integrate with Open WebUI, LibreChat, or AnythingLLM using the provided guides.

Доказательства

Piper has more than 170 voice files, some with hundreds of speakers. Their quality varies a lot, and many are licensed for non-commercial use only.

Источник

loudness-tts runs a rule layer first: money, dates, times, units, phone numbers, ranges, fractions, abbreviations and Roman numerals become words, in English and Spanish.

Источник

Every result is loudness-normalized to -16 LUFS with true peak at or below -1.5 dBTP, which suits podcasts and video.

Источник