Radar ·
Phonon: 40M-parameter multilingual TTS model released with speech-to-speech
Radar: the author announces Phonon, a text-to-speech model of about 40M parameters, multilingual, with a voice library and local use through the speech-to-speech package.
According to the author, Phonon is a text-to-speech release with a model of about 40M parameters covering several languages and a voice library. The source text states that it can be used locally through the speech-to-speech package, with the option --tts phonon. The associated description suggests that a small, locally runnable model may serve voice pipelines with low latency or on-device; this is the source's assessment, not a result measured here.
The source provides no quality metrics, comparisons, license or details beyond the date shown in Radar, 20 October 2026. To consult the original, find the announcement by its title on the author's page or the repository cited by your source, and check the license, version and supported languages before any use. Rota Nacional does not offer this model as a capability. Voice synthesis is in preview; audio transcription and document extraction already pass through the same personal data barrier.
Original source: Phonon: modelo TTS multilíngue de 40M de parâmetros lançado com speech-to-speech