Piper: A fast, local neural text-to-speech engine

Piper is a high-performance, local neural text-to-speech (TTS) engine designed for developers who need fast and private voice synthesis. It leverages espeak-ng for phonemization and ONNX Runtime for efficient neural network inference.

This project is ideal for applications where cloud-based TTS is not suitable due to privacy concerns, latency, or cost. By running entirely on-device, Piper provides a reliable and responsive voice synthesis solution.

Key Features

Fast & Efficient: Optimized for performance on a variety of hardware, from single-board computers like the Raspberry Pi to powerful desktop machines.
Fully Local: All processing happens on your device. No internet connection is required, ensuring complete privacy.
Wide Range of Voices: Supports numerous languages with a growing collection of high-quality voices available for download. See the full list in the Available Voices guide.
Multiple APIs: Interact with Piper through a simple Command-Line Interface, a Python API for easy integration, an HTTP server for network access, or a C/C++ API for high-performance applications.
Train Your Own Voices: The repository includes a complete toolchain for training new voices or fine-tuning existing ones on your own datasets. See the Training Guide.
Docker Support: A Dockerfile is provided for easy deployment and containerization. Check out the Docker Guide for more details.

Getting Started

Ready to get started? Head over to the Installation page to set up Piper, then follow the Quick Start guide for a simple "Hello World" example.

Community & Projects Using Piper

Piper is used by a variety of projects in the open-source community, including: