Skip to content
LiveMonitoring 39 sourcesעברית
AI·Newsroom

What’s new in AI: models, tools and techniques, verified

Open SourceReport1 min read

Hugging Face Details Phonon TTS and Introduces Modular Transformers Framework

Hugging Face highlights its open-source Phonon text-to-speech model and details a new modular framework designed to simplify adding models to Transformers.

AI Newsroom Automated DeskPublished Nov 7, 2026, 10:04 UTCPartially verified

In 30 seconds

Hugging Face has published details on Phonon TTS, an on-device text-to-speech model lineage featuring low latency and small parameter sizes. The organization also introduced a modular transformers approach intended to reduce code duplication when integrating new models.

Hugging Face Details Phonon TTS and Modular Architecture

Hugging Face has shared further details on its Phonon text-to-speech (TTS) model family, highlighting an open-source approach designed to run fully on user devices without relying on external servers.[1] The release builds on the Pocket TTS lineage, utilizing flow-matching techniques to generate audio and supporting multiple languages across compact parameter sizes.[1]

Alongside the voice model updates, Hugging Face detailed a new modular transformers approach for its Transformers library.[4] The framework allows developers to use imports and inheritance rather than traditional single-file configurations, significantly reducing the amount of code required to introduce new architectures.[4]

Phonon TTS Specifications and Capabilities

Phonon models are available in compact configurations around 65 MB to 75 MB in size, supporting languages including English, French, German, Spanish, and Portuguese.[1] According to the company, Phonon 31m is distributed under the Apache 2.0 license for free use in commercial applications, while other variants use non-commercial licensing.[1]

The accompanying ptts inference framework supports more than ten different hardware backends across CPUs, GPUs, and NPUs to optimize on-device execution.[1] Additionally, community conversions have brought Phonon 31M to run via WebGPU in browser environments using ExecuTorch.[3]

Evidence and Performance Notes

Performance figures and leaderboard standings for Phonon TTS are company-reported and have not been independently tested. At present, there is no independent reporting on this development.

Timeline

  1. Related publication by Hugging Face[2]
  2. Related publication by Hugging Face[3]
  3. Hugging Face publishes[1]
  4. First detected by AI Newsroom
  5. Verification completed

Sources

  1. [1]
  2. [2]
    Hugging FacePrimary sourceOct 2, 2026
    kyutai/pocket-tts · Hugging Face
  3. [3]
  4. [4]

How this article was produced

This article was researched and written by AI Newsroom’s automated editorial system using a large language model, constrained to cited evidence and checked by an automated quality gate. No human journalist wrote it.

This version was published automatically after passing quality checks, without human review.

AI Newsroom Automated Desk
Language model: gemini / gemini-3.5-flash-lite
Prompt version
article-draft-2026-11-07.2
Editorial standards

Related coverage

Open Source