Open-source licences

texttospeech.tools is built on the following third-party components. Each is used under the licence named here; the full licence texts are linked at the bottom of this page.

Voices and speech engine

Kokoro-82M v1.0 voice model and voice files, © hexgrad — Apache License 2.0. Source · ONNX export used here.

Text normalisation and phoneme post-processing adapted from kokoro-js 1.2.1, © hexgrad and contributors — Apache License 2.0. Source.

Transformers.js 3.8.1, © Hugging Face — Apache License 2.0. Source.

ONNX Runtime Web 1.22, © Microsoft Corporation — MIT License. Source.

Pronunciation (eSpeak NG)

Words are converted to pronunciations with eSpeak NG, compiled to WebAssembly by phonemizer.js 1.2.1. eSpeak NG is © its authors and licensed under the GNU General Public License, version 3 or later; phonemizer.js is © Joshua Lochner and licensed under the Apache License 2.0. This component is delivered to your browser as a separate file (the phonemizer script this site loads on demand) and is not combined with the rest of the site's code.

Corresponding source: eSpeak NG at github.com/espeak-ng/espeak-ng; the WebAssembly build scripts and wrapper at github.com/xenova/phonemizer.js (version 1.2.1). On request we will provide a copy of the exact corresponding source for the version served on this site, for at least three years from the last time we served it — write to support@texttospeech.tools.

You may replace or modify this component under the terms of the GPL; it is loaded as its own file and the rest of the site does not depend on its internals.

Audio encoding

The MP3 encoder includes the LAME MP3 encoder compiled to WebAssembly, © the LAME authors — GNU Lesser General Public License, version 2 or later. Source. It is loaded as a separate file only when you download an MP3, and can be replaced by the user under the terms of the LGPL.

Document import

Licence texts

Apache License 2.0 · MIT License · GNU GPL v3 · GNU LGPL v2 · Mozilla Public License 2.0