# Gradium Open-Sources Phonon, a 31M-Parameter On-Device TTS Model Under Apache 2.0

> **Publication**: aitalky (Independent AI Journalism)
> **Canonical Citation URL**: https://aitalky.vercel.app/news/gradium-open-sources-phonon-a-31m-parameter-on-device-tts-model-under-apache-2-0
> **Original Source**: [MarkTechPost](https://www.marktechpost.com/2026/10/22/gradium-phonon-31m-on-device-tts/)
> **Author**: Asif Razzaq (undefined)
> **Published Date**: Thursday, October 22, 2026
> **Category**: RESEARCH
> **Tags**: AI, MarkTechPost, research

## Executive Summary
Gradium has open-sourced Phonon, a family of on-device text-to-speech models. The English Phonon 31m ships under Apache 2.0, fits in 64.5 MB, and starts speaking in about 7 ms on a laptop CPU



## Full Coverage
Gradium has open-sourced Phonon, a family of on-device text-to-speech models. The English Phonon 31m ships under Apache 2.0, fits in 64.5 MB, and starts speaking in about 7 ms on a laptop CPU

---
*Generated for AI agents and LLM citation by aitalky (https://aitalky.vercel.app).*