MA Speech

A fully offline speech-to-text and text-to-speech app

Features:
- Offline speech recognition using whisper-base, bundled in the app
- Offline TTS using Supertonic 3, with ten voices across 31 languages
- Works as the system SpeechRecognizer and TTS service
- Every model ships inside the app, so it works on a clean install with no setup
- Runs on the GPU through a compute runtime built for this repo, not a vendor SDK

100% offline