Supertonic

Supertonic is a lightning-fast, on-device text-to-speech (TTS) system built for local inference using ONNX Runtime. It offers blazing-fast, real-time synthesis across various platforms (desktop, browser, mobile, edge) and supports 31 languages. Key features include a compact 99M-parameter open-weight model, enabling smaller downloads and faster startup times. Supertonic utilizes Expression Tags for nuanced speech and offers a multi-runtime SDK for seamless integration with languages like Python, Node.js, C++, Go, Swift, and Flutter. It outputs high-quality 44.1kHz WAV audio and includes a language-agnostic mode (“na”) for handling unknown input languages. The system is designed for edge device deployment, offering privacy and eliminating network dependencies. Demo and interactive examples are available, showcasing its capabilities, including handling complex expressions and phone numbers. Supertonic is ideal for applications needing local, private, and responsive TTS. The project utilizes Length-Aware RoPE and Self-Purifying Flow Matching for robust training. It's actively developed with support for multiple languages and runtime environments.

Key Features

• Lightning-fast on-device TTS
• Supports 31 languages
• 99M-parameter open-weight model
• Expression Tags for nuanced speech
• Multi-runtime SDK (Python, Node.js, C++, Go, Swift, Flutter)
• 44.1kHz High-Quality Audio Output
• Language-agnostic mode ('na')
• Edge device ready (zero network dependency)

Use Cases

• Creating audiobooks and podcasts for offline listening.
• Developing interactive voice assistants for mobile and embedded devices.
• Generating audio content for accessibility applications and assistive technologies.

Who is it for?

• Developers
• Copywriters
• Technical writers
• Marketing managers
• Product managers