Top AI Product

Supertone Inc shipped Supertonic v3 — 99M parameters, 31 languages, running entirely on-device via ONNX Runtime with zero cloud calls. GitHub trending lit up this week with 745+ daily stars as the broader dev community discovered the release.

## The size argument

At 99M parameters Supertonic v3 is roughly 7-20x smaller than competing open TTS systems (typically 0.7B to 2B). That translates to fast startup, small download, and viable smartphone or embedded deployment. Real-time synthesis works without a GPU. No API calls, no recurring costs, no privacy concerns — everything runs locally.

## What’s new in v3

31-language support, up from 5 in v2. No separate language adapters required — text can be processed language-agnostically. New expression tags: insert “, “, or “ directly in source text to control prosody. Reading-failure rate is meaningfully lower than v2 on long-form input.

## Why it matters

On-device TTS has been a frontier-lab privilege — most decent multilingual systems are either cloud-only or too heavy for prosumer hardware. Supertonic v3 combined with the Voice Builder workflow (released January 2026) closes that gap for indie developers, privacy-sensitive applications, and offline-first products. Weights available on Hugging Face under Supertone/supertonic-3.