Open-Source On-Device Text-to-Speech Model Phonon TTS Released
A Closer Look at Phonon TTS. Launching High-Quality On-Device Speech That Tops the Benchmarks.
Gradium has officially open-sourced Phonon TTS, an on-device text-to-speech model that runs entirely offline on smartphones, PCs, or single-board computers without requiring servers or internet access. Built on the Pocket TTS lineage and flow matching technology, the model is available in two sizes: a 31M version (approx. 65 MB, English only, commercially licensed under Apache 2.0) and a 40M version (approx. 75 MB, supporting five languages: English, French, German, Spanish, and Portuguese, under a CC BY-NC license).
The original post breaks down the engineering implementation of flow matching in ultra-compact speech models, providing a practical blueprint for deploying high-quality TTS on offline devices.
Source:Hugging Face · huggingface.co