Package: wnpp Severity: wishlist * Package name : piper1-gpl Version : 1.3.0 Upstream Contact: Michael Hansen <[email protected]> * URL : https://github.com/OHF-Voice/piper1-gpl * License : GPL-3.0 Programming Lang: C++, Python Description : A fast and local neural text-to-speech engine that embeds espeak-ng for phonemization.
Supports both training and inference and has a variety of voices in different languages. New voices can be trained and works with CPU or GPU. The model used is open and the data of voices that are shipped is available and uses various open licenses including public domain. Piper fills the gap between voices from festival and others that are used for accessibility with localization and multiple languages and more natural ones. Offers a Python package that installs voices as they are needed and it's possible to train new ones from some sample hours to clone voices locally. Does not requires huge amounts of space to store the models needed for TTS. Features a sample for various languages https://rhasspy.github.io/piper-samples/
