Voxify
Local, studio-grade neural text-to-speech desktop application powered by Kokoro-82M, ONNX Runtime, and Tauri v2.
Features
- Instant Speech & Selection Earcon: Under 2ms acoustic earcon chime and fast pipelined neural streaming.
- Authentic Human Prosody: Full sentence prosodic integrity preserved without robotic mechanical chunking.
- Deep Buffer Runway Pipeline: Pipelined background synthesis that maintains a continuous 6+ second buffer reservoir, eliminating playback pauses or underruns.
- Global Selection Reader: Select text in any desktop app (browser, PDF reader, editor, Word) and Voxify speaks immediately.
- Comprehensive Text Normalization: Handles honorifics, middle initials, Latin citations, units of measurement, dialogue attribution tags, markdown links, and HTML tags effortlessly.
Architecture Documentation
For the complete technical specification, core invariants, abbreviation registry, clause splitting algorithm, and testing guide, see:
Development & Verification
To run the full automated verification test suite:
cd src-tauri
cargo test --lib -- --nocapture
To run the application locally:
.\start-voxify.bat
# Or
npm run tauri dev