Repolode
Submit
OpenMOSS

OpenMOSS/MOSS-TTS-Nano

A 0.1B-parameter multilingual text-to-speech model that runs in realtime on CPU, with voice cloning and 48 kHz stereo output.

Open-source alternative to:ElevenLabs

View on GitHubWebsite
Stars
4,517
Forks
562
Open issues
26
Last commit
last month

What is it?

MOSS-TTS-Nano is a small text-to-speech model from the OpenMOSS team, with about 100M parameters. It is aimed at local and lightweight deployments where a GPU is not available, and it generates speech in realtime on CPU using an autoregressive audio tokenizer plus LLM pipeline.

The model supports voice cloning from a reference audio clip and outputs 48 kHz, two-channel audio. It covers 20 languages, including Chinese, English, Japanese, Korean and several European languages. Long text can be split into chunks automatically for cloning.

The project provides several entry points: a Python inference script, a local FastAPI web demo, a packaged CLI with generate and serve commands, and an ONNX Runtime CPU backend that does not need PyTorch at inference time. Fine-tuning code and an Android ONNX example are also included.

Key features

  • 01Roughly 0.1B parameters, designed to run on a 4-core CPU with streaming output.
  • 02Voice cloning from reference audio, with automatic chunking for long text.
  • 03Outputs 48 kHz two-channel audio.
  • 04Supports 20 languages including Chinese and English.
  • 05ONNX Runtime CPU backend with no PyTorch dependency during inference.
  • 06Includes a web demo, CLI, and finetuning code.

Use cases

  • Running a local text-to-speech service on a laptop or CPU-only server.
  • Adding voice output with cloned voices to a lightweight product or prototype.
  • Building a browser or offline reading tool that generates speech locally.
  • Fine-tuning a small multilingual TTS model on custom voice data.

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/OpenMOSS/MOSS-TTS-Nano.git
  2. 2

    Enter the project folder and install the requirements

    pip install -r requirements.txt
  3. 3

    Install the project in editable mode

    pip install -e .
  4. 4

    Run voice cloning inference with the sample prompt audio

    python infer.py --prompt-audio-path assets/audio/zh_1.wav --text "欢迎关注模思智能、上海创智学院与复旦大学自然语言处理实验室。"

Things to know

  • Installing dependencies such as pynini and WeTextProcessing can fail and may need manual workarounds.
  • The README's Conda and pip setup steps are more involved than a single package install.

Topics

#text-to-speech#voice-cloning#cpu-inference#onnx#multilingual#realtime#speech-synthesis#audio-tokenizer#chinese#english

FAQ

What is MOSS-TTS-Nano?
A 0.1B-parameter multilingual text-to-speech model that runs in realtime on CPU, with voice cloning and 48 kHz stereo output. MOSS-TTS-Nano is a small text-to-speech model from the OpenMOSS team, with about 100M parameters. It is aimed at local and lightweight deployments where a GPU is not available, and it generates speech in realtime on CPU using an autoregressive audio tokenizer plus LLM pipeline.
Is MOSS-TTS-Nano free to use? What license does it have?
MOSS-TTS-Nano is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is MOSS-TTS-Nano still maintained?
Updated within 6 months. The last commit to MOSS-TTS-Nano was on Sep 6, 2026.
What can I use MOSS-TTS-Nano for?
Running a local text-to-speech service on a laptop or CPU-only server. Adding voice output with cloned voices to a lightweight product or prototype. Building a browser or offline reading tool that generates speech locally. Fine-tuning a small multilingual TTS model on custom voice data.
How do I install MOSS-TTS-Nano?
Clone the repository: git clone https://github.com/OpenMOSS/MOSS-TTS-Nano.git Enter the project folder and install the requirements: pip install -r requirements.txt Install the project in editable mode: pip install -e . Run voice cloning inference with the sample prompt audio: python infer.py --prompt-audio-path assets/audio/zh_1.wav --text "欢迎关注模思智能、上海创智学院与复旦大学自然语言处理实验室。"
Where can I get MOSS-TTS-Nano?
The source code and installation instructions are on GitHub: https://github.com/OpenMOSS/MOSS-TTS-Nano

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects