Third-Party Licenses

Runtime dependencies

Package	License	Role
phonemize (hans00)	MIT	Default G2P phonemizer (hans00 fork)
onnxruntime-react-native	MIT	Neural inference (optional peer)
@dr.pogodin/react-native-fs	MIT	File I/O (peer)

Library data

This library ships no dictionary or model data. Consumer apps supply these at runtime:

• Phonemization dictionary (optional): apps may load any EPD1-format mmap dict via the NativeDict API. The library default phonemizer is phonemize (MIT, hans00 G2P) which requires no dict. See PHONEMIZATION.md for building and attributing a dict.

Neural TTS model licenses

Models are NOT bundled. Each engine loads models from paths provided by the consumer app. Upstream model licenses differ — review before integrating:

Engine	Upstream source	Code license	Model weights license	Notes
Kokoro	https://huggingface.co/hexgrad/Kokoro-82M	Apache-2.0	Apache-2.0	Fully permissive, commercial use allowed.
Supertonic	https://huggingface.co/Supertone/supertonic (v1), https://huggingface.co/Supertone/supertonic-2 (v2)	MIT (repo)	OpenRAIL	Reference code is MIT; the model weights are under an OpenRAIL-style responsible-AI license (use-based restrictions, no absolute commercial ban). Review the model card before shipping.
Kitten	https://huggingface.co/KittenML (upstream); example app uses palshub/* mirrors	Apache-2.0	Apache-2.0	Fully permissive, commercial use allowed.

Example-app download URLs (example/src/utils/SupertonicModelManager.ts, example/src/utils/KittenModelManager.ts) point at those upstream repos or author-owned mirrors; the library itself bundles no weights.

Consumer apps are responsible for complying with each upstream model's terms of use. MIT-licensed library code does not inherit the model's license — comparable to VLC (GPL) playing CC-BY-NC video: the runtime and the content are independently licensed.

This library

MIT. See LICENSE.