From the source
Gradium has published a closer look at Phonon, the small on-device text-to-speech model it opened up two weeks ago, covering its design, demos and leaderboard results.
- Phonon runs on the user's own device with no server and no network, so the text typed into it stays local rather than making a round trip to a cloud service.
- The models come in two sizes, about 65 MB and 75 MB, and together speak English, French, German, Spanish and Portuguese, with more than 100 built-in voices and custom voices made in Gradium Studio.
- Phonon 31m speaks English under Apache 2.0, while Phonon 40m covers all five languages under a non-commercial license; commercial use of the latter is priced by number of users, not by speech generated.
- On
Hugging Face's Open TTS Leaderboard, Gradium says Phonon ranks first in German and Spanish for intelligibility and in the top three in every language evaluated, next to models up to 100 times its size.
- Gradium also built ptts, an inference framework supporting more than 10 backends across CPU, GPU and NPU, and Phonon is available in Transformers from v5.21 as its reference PyTorch implementation.






