We're developing our own implementations of text-to-speech and speech-to-text to use in #GrapheneOS which are entirely open source and avoid using so-called 'open' models without the training data ava...