VoxNemesis Supertonic

The Challenge

High-quality text-to-speech usually requires cloud APIs, sending user data to third-party servers. Users needed a private, fast, and truly local alternative that runs entirely in the browser without compromising on audio quality or latency.

The Solution

VoxNemesis Supertonic is a Chrome Extension (Manifest V3) that runs the Kokoro TTS model directly in the browser using ONNX Runtime Web. It achieves sub-second synthesis times with WebGPU acceleration, while falling back gracefully to WASM for compatibility. No data ever leaves the user's machine.

Technical Implementation

The extension uses the Offscreen API to create a dedicated audio engine that persists even when the popup closes. It implements a watchdog pattern to handle race conditions between the popup and background worker during initial model download. The architecture includes intelligent backend switching — attempting WebGPU first, then gracefully degrading to WASM with SIMD for broader hardware compatibility.