OpenMOSS/MOSS-TTS-Nano

MOSS-TTS-Nano is an open-source multilingual tiny speech generation model from MOSI.AI and the OpenMOSS team. With only 0.1B parameters, it is designed for realtime speech generation, can run directly on CPU without a GPU, and keeps the deployment stack simple enough for local demos, web serving, and lightweight product integration.

What it solves

MOSS-TTS-Nano is designed to make high-quality, multilingual speech generation accessible for real-time applications. It addresses the challenge of deploying text-to-speech (TTS) systems that are typically resource-heavy, by providing a model with a tiny footprint and low latency that can run efficiently on standard CPUs without requiring a GPU.

How it works

The project uses a pure autoregressive pipeline consisting of an Audio Tokenizer and a Large Language Model (LLM). Specifically, it employs MOSS-Audio-Tokenizer-Nano—a lightweight, CNN-free tokenizer based on causal Transformer blocks—to compress 48 kHz stereo audio into a token stream. The LLM then generates these tokens to produce speech. For deployment, it offers an ONNX CPU version that removes the PyTorch dependency, significantly increasing processing efficiency and allowing it to run on a single CPU core.

Who it’s for

  • Developers looking to integrate lightweight TTS into local demos, web services, or mobile applications (including Android).
  • Product teams needing real-time, multilingual speech generation with minimal infrastructure costs.
  • Researchers interested in tiny, efficient autoregressive audio models.

Highlights

  • Tiny Model Size: Only 0.1B parameters.
  • CPU-Friendly: Capable of streaming generation on a 4-core CPU and smooth inference on a single core via ONNX.
  • Multilingual Support: Supports 20 languages, including Chinese, English, German, Spanish, and French.
  • High Fidelity: Native 48 kHz, 2-channel (stereo) audio output.
  • Voice Cloning: Supports voice cloning using a reference audio prompt.
  • Flexible Deployment: Provides a packaged CLI, FastAPI web demo, and ONNX runtime support.

Related

  • Project
    Audio8-AI/Audio8_TTS

    A compact 0.6B-parameter multilingual text-to-speech model featuring zero-shot voice cloning and efficient deployment options for CPU and GPU.

  • Project
    samuel-vitorino/sopro

    A lightweight 120M-parameter text-to-speech model family that enables high-quality voice cloning and streaming audio generation on laptop CPUs or in the browser.

  • Project
    thewh1teagle/kokoro-onnx

    A lightweight ONNX Runtime implementation of Kokoro-TTS that enables fast, multi-language text-to-speech generation on CPU and GPU.

  • Project
    KittenML/KittenTTS

    A lightweight, ONNX-based text-to-speech library that provides high-quality voice synthesis on CPU with models as small as 25MB.

  • Project
    tronghieuit/v-tts

    A lightweight Vietnamese text-to-speech system featuring multi-speaker synthesis and zero-shot voice cloning that runs efficiently on CPU.