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

Gradium has published a look back at the launch of Phonon, a compact text-to-speech model that runs entirely on a user's device, with no server or network connection. In a Hugging Face blog post dated November 5, 2026, the company says the model was downloaded more than 200,000 times in its first two weeks.

Phonon comes in two sizes of roughly 65 MB and 75 MB and offers more than 100 built-in voices. Phonon 31m speaks English and carries an Apache 2.0 license that permits commercial use. Phonon 40m adds French, German, Spanish and Portuguese but is licensed for non-commercial use only, with commercial pricing based on user counts. Gradium says it runs on phones, laptops, single-board computers and in browsers, and that it builds on the Pocket TTS lineage, using flow matching, a method that proved numerically unstable early in training.

On Hugging Face's Open TTS Leaderboard, Gradium reports that Phonon ranked first for intelligibility in German and Spanish and placed in the top three in every language it evaluated, against models up to 100 times larger. The company also says it leads the CPU streaming benchmark for time-to-first audio.

Developers can use a new inference framework, ptts, with more than 10 CPU, GPU and NPU backends, or Hugging Face Transformers from version 5.21. The release matters because it pairs competitive measured quality with a footprint small enough for private, offline voice features. Gradium also offers custom training for design partners, including smaller models and languages such as Korean, Arabic or Japanese.