Supertonic MNN

English | 中文

---

English

Supertonic MNN is a high-performance, lightweight text-to-speech (TTS) library based on MNN. It supports both command-line interface (CLI) and Python API, making it easy to integrate into your projects.

Demo video: https://www.bilibili.com/video/BV1VFqiBSER3

Features

• Fast Inference: RTF ~ 0.07 on CPU.
• Lightweight: Minimal dependencies.
• Supported Precisions: fp32, fp16, int8.

Available Voices

• M1: Male voice 1
• M2: Male voice 2
• F1: Female voice 1
• F2: Female voice 2

Documentation

Full documentation is available at Supertonic MNN Docs.

Installation

pip install supertonic-mnn

Quick Usage

CLI

echo "Hello world" | supertonic-mnn -o hello.wav

Python API

from supertonic_mnn import SupertonicTTS

# 1. Initialize
tts = SupertonicTTS()

# 2. Synthesize
# Models will be downloaded automatically if not present
audio, sample_rate = tts.synthesize("Hello world", voice="M1", output_file="hello.wav")

See examples/ for more details.

Acknowledgments

This project is based on the original Supertonic by Supertone Inc.

---

中文

Supertonic MNN 是一个基于 MNN 的高性能、轻量级文本转语音 (TTS) 库。它同时支持命令行接口 (CLI) 和 Python API，方便您将其集成到项目中。

Demo video: https://www.bilibili.com/video/BV1VFqiBSER3

特性

• 极速推理: CPU 上 RTF 约为 0.07。
• 轻量级: 依赖极少。
• 多精度支持: fp32, fp16, int8。

可用音色

• M1: 男声 1
• M2: 男声 2
• F1: 女声 1
• F2: 女声 2

文档

完整文档请访问 Supertonic MNN 文档。

安装

pip install supertonic-mnn

快速上手

命令行 (CLI)

echo "你好，世界" | supertonic-mnn -o hello.wav

Python API

from supertonic_mnn import SupertonicTTS

# 1. 初始化
tts = SupertonicTTS()

# 2. 推理
# 如果模型不存在，会自动下载
audio, sample_rate = tts.synthesize("你好，世界", voice="M1", output_file="hello.wav")

更多详情请参阅 examples/ 目录。

致谢

本项目基于 Supertone Inc. 的原始 Supertonic 工作。