Open-source licences
Atmos1 uses programs and voices made by other people. This page names them and says what licence each one comes with. The same notices are inside the app.
Piper, the natural voices' program (Windows)
Piper turns text into speech for the natural voices you download. It runs on your computer.
- Piper. MIT licence. Copyright (c) 2022 Michael Hansen. github.com/rhasspy/piper
- piper-phonemize. MIT licence. Copyright (c) 2023 Michael Hansen. github.com/rhasspy/piper-phonemize
- ONNX Runtime. MIT licence. Copyright (c) Microsoft Corporation. github.com/microsoft/onnxruntime
- espeak-ng. GNU General Public License, version 3 (GPL-3.0). It knows how to say words. github.com/rhasspy/espeak-ng and github.com/espeak-ng/espeak-ng
The full text of each licence is in the Atmos1 app folder, in "licenses", and is also on the pages above.
Offer of the source code of espeak-ng
espeak-ng is free software, licensed under the GNU General Public License, version 3. Atmos1 runs it as a separate program and does not change it. You may have the complete source code of espeak-ng, and of the piper-phonemize library that uses it, for the version in your copy of Atmos1. It is free, apart from the cost of copying.
To ask, use Report a problem in Atmos1, or write to us using the contact on the privacy notice, and say which version of Atmos1 you have. This offer lasts three years from the date of that version. The same source is public at the addresses above.
The natural voices
The voices are not part of the app. Atmos1 downloads the ones you ask for from huggingface.co/rhasspy/piper-voices, at one fixed version, and checks each file before it uses it. Each voice keeps its own licence. These are the voices Atmos1 offers:
- Cori (British, medium and high). Recordings by volunteer readers from LibriVox. Public domain.
- Kristin (American). Recordings by volunteer readers from LibriVox. Public domain.
- John (American). Recordings by volunteer readers from LibriVox. Public domain.
- Bryce (American). Recordings by volunteer readers from LibriVox. Public domain.
- Norman (American). Recordings by volunteer readers from LibriVox. Public domain.
- Linda (American, medium and high). The LJ Speech Dataset, by Keith Ito and Linda Johnson. Public domain.
The voice models were made with the Piper training tools by the Piper project and its contributors.
Other software in Atmos1
Atmos1 also uses whisper.cpp and OpenAI Whisper models for turning speech into text on your computer, Electron, and a small number of open-source libraries. Their names, versions and licences are in the notices file inside the app.
← Back to Atmos1