The development

What happened

Gradium’s 5 November technical walkthrough describes Phonon, an on-device text-to-speech family opened two weeks earlier. The post explains variant licensing and inference options; it is not an announcement that the models first became available today.

Key facts

  • The post identifies English-only Phonon 31m as Apache 2.0, while Phonon 40m covers English, French, German, Spanish and Portuguese under non-commercial licensing.

  • Gradium describes ptts as an on-device, single-stream inference framework with CPU, GPU and NPU backends. Transformers support from version 5.21 targets batched GPU generation.

  • The authors report leaderboard results, but these are not independently reproduced here. Availability of local inference does not establish the privacy behavior of every application or voice-creation service.

The context

Why it matters

Developers evaluating local speech need to distinguish variant permissions and deployment paths, rather than assuming all open weights have the same terms. The walkthrough supports a practical integration comparison, not a universal quality or privacy guarantee.