Phonon: open 31M-parameter TTS running in the browser via WebGPU
Post reports a port of Phonon, an open 31M-parameter text-to-speech model from GradiumAI, running fully in the browser via WebGPU in half precision, with the author citing about 47x real time on an M4 Pro and 7x on a Phone 13 Pro.
Source: Phonon: TTS aberto de 31M de parâmetros rodando no navegador via WebGPU (x.com). Text prepared with AI from this source.
What happened and what to do
The post reports a port of Phonon, an open text-to-speech model with 31M parameters developed by GradiumAI, to run entirely in the browser via WebGPU in half precision. According to the author, the model reaches about 47x real time on an M4 Pro and 7x on a Phone 13 Pro, indicating on-client speech synthesis on common hardware.
For companies, the practical takeaway is that voice features such as content reading, guided support and accessibility could run on the user's own device without sending text or audio to an external service. This reduces per-call cost, network latency and exposure of sensitive data. It is worth measuring quality and performance with the real voices, languages and devices of your audience before choosing between local synthesis, a private server or a hybrid model.
How the consultancy can help
Wendelmaques can diagnose where on-device voice makes sense for your product, measuring quality, latency and consumption on real devices, and implement the TTS layer in the browser or app, with a private-server fallback when needed.
Next step
Send a short description of your voice use case, including channels, usage volume and devices served, and receive a scoped proposal from Wendelmaques.
Let's assess your project
I reply with scope and a proposal.