Piper
Piper is a fast, local neural text-to-speech engine maintained by the Open Home Foundation and used by Home Assistant, the NVDA screen reader and LocalAI. It embeds espeak-ng for pronunciation, installs with pip, and offers a command line, a web server, a Python API and a C library. The engine is GPL-3.0; each voice model carries its own licence.
What does Piper do?
Piper is "a fast and local neural text-to-speech engine that embeds espeak-ng for phonemization", in its README's words. It is built for speed on small hardware rather than for cloning a voice: you pick a trained voice model and it speaks.
The project offers a command-line interface, a web server, a Python API, a C and C++ library (libpiper), a Java binding, and documentation for training your own voice. The README names Home Assistant, the NVDA screen reader, LocalAI and Open Voice Operating System among the projects that use it.
Key facts
- Licence: GPL-3.0 on the engine (OHF-Voice/piper1-gpl). The archived original at rhasspy/piper was MIT and its README redirects here.
- Embeds espeak-ng for phonemization, which is how it turns text into sounds before the neural model speaks them.
- Interfaces documented in the README: command line, web server, Python API, C and C++ library (libpiper), Java binding, and voice training.
- Named users in the README: Home Assistant, the NVDA screen reader, LocalAI, and Open Voice Operating System.
- Voice samples are published at rhasspy.github.io/piper-samples; voice licences are listed in the project's separate voices document.
- Latest tagged release when read: v1.8.0 on 2026-09-04 (GitHub releases API).
- The README carries the note that the Open Home Foundation is looking for maintainers.
What does it replace, and where does it fall short?
Piper replaces ElevenLabs where the job is to speak text on a device or a home server with no network call and no account, and where a preset voice is enough. The engine code is GPL-3.0, so a product that links it has to honour that licence; the voice models each carry their own licence, listed in the project's separate voices document, which this page has not read.
Where it falls short: the README does not list languages, does not describe voice cloning, and does not state a sample rate or streaming support. The samples page linked above is the way to hear what a given voice sounds like.
How does Piper run?
pip install piper-tts for the CLI and the Python API. A web server mode exists for HTTP. The C/C++ library is for embedding. Hardware requirements are not stated, and the projects that use it, such as Home Assistant, run it on small machines.
The original repository at rhasspy/piper is archived and its README points to this one. The old repository was MIT; the current one is GPL-3.0.
Who is Piper for?
A team building something that runs at the edge or in a home, a screen reader, or an assistant that must keep working offline. A team that wants voice cloning or an OpenAI-compatible API should look at the other entries; Speaches, under speech-to-text engines, serves Piper voices through that API shape.
What limits does the README state?
"The Open Home Foundation is looking for maintainers for Piper!" That is the project's own note on its capacity.
Questions people ask
Is Piper open source?
Yes. The engine is under GPL-3.0, which is an OSI approved licence. The voice models are downloaded separately and each carries its own licence, listed in the project's voices document, so a product should check the licence of the specific voice it ships, not only the engine.
Can Piper be used commercially?
The GPL-3.0 engine can be used commercially, and a product that links it has to honour the GPL's terms on distributing source. The voices are the second check: their licences vary by voice. Home Assistant and NVDA, both named in the README, ship it.
Does Piper run offline and on small devices?
Yes. It is built for local use with no network call, and its README describes it as fast and local. Home Assistant runs it on home servers and small machines. Hardware minimums are not stated in the README.
How does Piper compare with ElevenLabs?
Piper speaks with a preset trained voice on a device with nothing sent anywhere; ElevenLabs clones voices and claims 70+ languages from a hosted API. Piper does not clone and its README does not list languages. The open-source alternatives to ElevenLabs page places it as the pick for small devices and offline assistants.
Where did the original Piper repository go?
Development moved from rhasspy/piper, which is archived, to OHF-Voice/piper1-gpl under the Open Home Foundation. The old repository's README is one line pointing to the new one. The licence changed in the move, from MIT to GPL-3.0.
Sources
- Piper README and LICENSE: github.com/OHF-Voice/piper1-gpl, read 2026-09-16.
- The archived original: github.com/rhasspy/piper, read 2026-09-16.
Compared with the others
On the open-source alternatives to ElevenLabs page, Piper is the pick for small devices and offline assistants. The fast local engine behind Home Assistant and the NVDA screen reader, with a pip install, a web server and a C library.
Also on that page: Chatterbox for voice cloning you can ship, Kokoro FastAPI for a speech api on your own server, Fish Speech for expressive cloning, non-commercial.
More text to speech and voice cloning
An AGPL-3.0 desktop workflow engine for voice cloning, voice design, video dubbing, dictation and audiobooks, running 16 speech engines and 10 transcription engines locally, with a local API and an MCP server for agents.
A multilingual text-to-speech system with voice cloning from a 10 to 30 second clip and inline emotion tags, released with its weights under a research licence that requires a separate agreement for commercial use.
A family of text-to-speech models from Resemble AI under MIT, with zero-shot voice cloning from a reference clip and a multilingual model covering 23 languages.
A research text-to-speech model with zero-shot voice cloning whose code is MIT but whose pretrained weights are non-commercial, because of the dataset they were trained on.
A Docker image that serves the 82-million-parameter Kokoro model through an OpenAI-compatible speech endpoint, with streaming, voice mixing and nine languages, all under Apache-2.0.
Tencent Hunyuan's 1.5-billion-parameter speech foundation model under MIT for code and weights, covering zero-shot text to speech, speech editing, enhancement and separation through one natural-language instruction interface.
Added September 16, 2026. Every claim above comes from the project's README, LICENSE or model card, read on September 16, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/OHF-Voice/piper1-gpl.
Want Piper running on your own servers?
Reveneau sets it up on your hardware, connects it to your product, and builds the software around it.