Favicon of MOSS TTS

MOSS TTS

An open-source text-to-speech model for multilingual narration and voice cloning, with pronunciation and timing controls, licensed under Apache 2.0.

Screenshot of MOSS TTS website

MOSS TTS generates speech for developers and creators working on narration, dubbing and other long audio content. It supports consistent voices across long passages and can clone a speaker from reference audio without separate training for that voice.

  • Voice cloning and continuation: Generate speech in a reference speaker's voice or continue an existing audio recording.
  • Multilingual narration: Produce speech in languages including English, Chinese, Cantonese, Japanese, Hindi and Arabic, and mix languages within a passage.
  • Pronunciation and pacing: Use Pinyin or phonemes to control pronunciation and adjust duration and pauses for scripts that need precise delivery.
  • Stereo audio: The MOSS-TTS-Local-Transformer variant supports native stereo input and output and streaming audio decoding.

MOSS TTS is an open-source model under Apache 2.0, which permits commercial use. It provides downloadable models for deployment on your own hardware through Hugging Face and ModelScope.

Similar to MOSS TTS