Open source licenses
Axo is built on open source work. The components below are included in the app or downloaded by it, with our thanks to their authors.
Supertonic (Supertone Inc.)
On-device text-to-speech. The inference code is a port of Supertone's official
Swift reference, used under the MIT License; the model weights are
used under the OpenRAIL-M License, which permits commercial use and
attaches use restrictions aimed at preventing harm — including a prohibition on
generating speech intended to deceive or to impersonate a person without consent.
Axo uses the model only to voice its own fictional tutor characters.
github.com/supertone-inc/supertonic
Kokoro-82M
Text-to-speech model and tokenizer, Apache License 2.0.
huggingface.co/hexgrad/Kokoro-82M
·
ONNX build
Phonemis (Igor Swat)
Grapheme-to-phoneme engine, MIT License.
github.com/IgorSwat/Phonemis
xsimd (xtensor-stack)
SIMD primitives used by the phoneme engine, BSD 3-Clause License.
github.com/xtensor-stack/xsimd
ONNX Runtime (Microsoft)
Neural network inference engine, MIT License.
github.com/microsoft/onnxruntime
React Native and Expo
Application framework, MIT License.
github.com/facebook/react-native
·
github.com/expo/expo
Everything else
Axo also depends on a number of smaller open source packages, each under its own permissive license (MIT, BSD, or Apache 2.0). For a full dependency list or a copy of any license text, write to brucegarro@gmail.com.