A Closer Look at Phonon TTS. Launching High-Quality On-Device Speech That Tops the Benchmarks.
huggingface ·
Gradium has released Phonon, a high-quality text-to-speech model designed to run entirely on local devices without requiring internet connectivity or cloud servers. The models are available in two small sizes, approximately 65 MB and 75 MB, supporting English, French, German, Spanish, and Portuguese with over 100 built-in voices. Phonon utilizes flow matching techniques derived from the Pocket TTS lineage to achieve superior audio quality despite its compact footprint. In its first two weeks, the project garnered significant attention, recording over 200,000 downloads and trending on Hugging Face. On the Hugging Face Open TTS Leaderboard, Phonon ranked first in intelligibility for German and Spanish while topping CPU streaming benchmarks for speed. The release includes an Apache 2.0 licensed version for commercial use and non-commercial variants, alongside a new inference framework called ptts that supports various hardware backends like CPUs, GPUs, and NPUs. Developers can integrate Phonon into applications using the Transformers library or custom implementations, enabling features like voice cloning and offline operation.
huggingface