Piper Download: Fast, Local Neural Text-to-Speech

Piper reads text aloud with neural voices, quickly and entirely on your own device. It is used by projects such as Home Assistant, the NVDA screen reader and Open Voice OS. The original repository (rhasspy/piper, MIT-licensed) was archived on October 6, 2025, and development moved to OHF-Voice/piper1-gpl under the Open Home Foundation; that is the version described here. The foundation has also said it is looking for maintainers.

What does Piper do?

• Offline speech: Generates speech locally, with no cloud service. It uses espeak-ng for phonemization.
• Many voices: Downloadable voice models for a range of languages and speakers.
• Several interfaces: A command-line tool, an HTTP web server, a Python API and a C/C++ library (libpiper).
• Training: Documentation for training your own voices.

Where Piper fits

It is ideal for smart-home setups, accessibility tools, voice assistants and developers who need fast, private text-to-speech.

How to install

Piper is published on PyPI and needs Python 3.9 or newer:

pip install piper-tts

First steps

List the available voices and download one (it is saved in the current folder unless you pass --data-dir):

python3 -m piper.download_voices
python3 -m piper.download_voices en_US-lessac-medium

Turn text into a WAV file:

python3 -m piper -m en_US-lessac-medium -f test.wav -- 'This is a test.'

Leave out -f to play the speech directly (this needs ffplay from ffmpeg). The command line reloads the voice each time, so for repeated use the project recommends its web server.

System requirements

• Systems: Ready-made packages exist for Windows (64-bit), macOS (Intel and Apple Silicon) and Linux (x86_64 and ARM64).
• Python: 3.9 or newer.
• Hardware: No GPU is needed. The optional --cuda flag enables GPU acceleration with the onnxruntime-gpu package.

License

The current Piper (piper1-gpl) is free and open source under the GPL-3.0 license. Each voice model has its own license.