A Closer Look at Phonon TTS. Launching High-Quality On-Device Speech That Tops the Benchmarks.
Web page · https://huggingface.co/blog/phonon-tts
First seen by Oriel on .
AI summary
Gradium’s Phonon is a compact, open-weight text-to-speech model designed to run offline on-device, with strong leaderboard results and support for multiple inference backends.
- Two model sizes, about 65 MB and 75 MB, cover English, French, German, Spanish, and Portuguese, with more than 100 built-in voices.
- Phonon 31m is English-only under Apache 2.0; Phonon 40m supports all five languages under a non-commercial license.
- It ranks first in German and Spanish intelligibility and in the top three for every evaluated language; its models also top the CPU streaming benchmark for time-to-first audio.
- The ptts inference framework supports more than 10 CPU, GPU, and NPU backends, while the Transformers implementation targets batched GPU generation and model development.
- Users can create custom voices through voice cloning or voice design, and the release received more than 200,000 downloads in its first two weeks.
#text-to-speech #on-device inference #transformers #voice cloning
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Blog | https://huggingface.co/blog/phonon-tts | https://huggingface.co/blog/phonon-tts Read discussion |