Licenses and credits.

Bostaji's voice

The voices that read your emails and answer you are synthesized entirely on your machine. They are built on models published by the community:

• Jessica and Pierre voices — models built from the UPMC dataset (Université Pierre-et-Marie-Curie), published under the CC-BY-SA 4.0 license. The listening samples bundled with the application, produced with these voices, are shared under the same license.
• Fallback voice — Kokoro model (Apache 2.0 license), voice built from the SIWIS dataset, published under the CC-BY 4.0 license.

Text-to-speech engine

Local speech synthesis relies on the Piper engine (the piper-tts package), published under the GPL-3.0 license by the Home Assistant authors. It is distributed unmodified alongside the application, in a separate program; its source code is available at github.com/OHF-voice/piper1-gpl.