DigestAI news desk

Cut through the AI noise.

Hardware & Computeupdated 6 min read

Hugging Face releases Phonon TTS with on-device speech and open weights

Hugging Face launched Phonon, a text-to-speech model designed to run entirely on a user’s device without cloud dependency. The release includes two models: Phonon 31m (Apache 2.0 license, supports English) and Phonon 40m (CC BY-NC, supports five languages: English, French, German, Spanish, and Portuguese). Both are under 75 MB and feature 100+ built-in voices, with custom voice cloning via…

2 sources primary source

Key points

  • Phonon runs offline on devices like phones, laptops, and browsers with no cloud dependency
  • Two models released: **Phonon 31m** (English, Apache 2.0) and **Phonon 40m** (5 languages, CC BY-NC)
  • Tops benchmarks for intelligibility and speed, supports 100+ voices and custom cloning

The model tops benchmarks on intelligibility (first in German/Spanish, top 3 across languages) and speed (fastest CPU streaming). Hugging Face built a new inference framework (ptts) for cross-device compatibility, including support for CPU, GPU, and NPU backends. Phonon integrates with Hugging Face Transformers (v5.21+) and has been downloaded 200,000+ times in its first two weeks. The company also introduced the Open TTS Leaderboard, a new tool for comparing open TTS models across 23 languages.

Model page: Phonon 31m →

The story so far

2 episodes →
  1. Hugging Face releases Phonon TTS with on-device speech and open weightsthis story
Full story from Hugging Face · by Olesia Krizhevich primary sourceOpen source ↗

A Closer Look at Phonon TTS. Launching High-Quality On-Device Speech That Tops the Benchmarks.

Hugging Face · 5 November 2026

Loading the full article…

This text was published by Hugging Face and written by Olesia Krizhevich. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗

Coverage and discussion

2sources
Topics · follow one to build your own front page

The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.

Comments

via GitHub Discussions

More in Hardware & Compute

All →

Related stories