Piper is a fast, local neural text-to-speech engine that converts text into spoken audio on your own hardware, with no cloud service involved. It embeds espeak-ng for phonemization and installs with a single pip command. The project provides a command-line interface, a web server, a Python API and a C/C++ API, plus guidance for training new voices and building manually. This repository is maintained under the Open Home Foundation, is licensed GPL-3.0, and is free to use.
Piper is aimed at developers, hobbyists and accessibility projects that want offline voices with low latency. Smart home users get spoken responses through Home Assistant, and screen reader users rely on it through NVDA integrations. Engineers building voice assistants on small devices such as Jetson boards can run it locally, and teams building local AI stacks use it through projects like LocalAI. Programmers working in other languages can use bindings, including a Java option called Piper JNI. Researchers can train their own voices using the documented training workflow, and the Open Home Foundation is currently looking for new maintainers.
Typical use cases include reading Home Assistant notifications aloud, powering an offline voice assistant that never sends audio to a third party, and generating narration for videos and e-learning with scripted command-line calls. Developers expose a text-to-speech endpoint through the built-in web server for internal apps, or call the Python API to synthesize audio inside a program. Accessibility tools read clipboard content aloud, including with language detection, and game plugins add spoken dialogue. Because everything runs locally, it also suits privacy-sensitive deployments and environments without reliable internet access.
