Guide

How on-device (browser) text to speech works

Last updated: 23 October 2026

Most text-to-speech services send your words to a company's servers, generate the audio there, and send it back. On-device text to speech does it differently: the speech model runs inside your own browser, so your text is turned into audio locally and never leaves your device. This guide explains how that works and what it means for you.

A model that runs in your browser

YourAIVoice downloads an open-source voice model to your browser once (about 95 MB) and keeps it cached. From then on, your device does the work of turning text into speech using WebAssembly, the same technology that lets demanding software run in a browser. There is no server step for the everyday voices.

Why it is private

Because the text is processed on your device, it is never uploaded or stored anywhere. The audio and captions exist only in your browser until you download them. That is a stronger privacy guarantee than a policy promise — the data simply never leaves.

Why it works offline

Once the model is cached, no connection is needed to generate speech, so the installed app works on a flight or anywhere with no internet. See offline text to speech.

The trade-offs

On-device speech means a one-time download and relies on your device's processor, so very long text takes a little longer than a cloud service with a data centre behind it. Some languages are also harder to cover well on-device — for example, the on-device Telugu voice is private but synthetic, so YourAIVoice offers an optional natural Telugu voice via Microsoft Azure, which does send text to Microsoft. The choice is always clearly labelled.

Try it

Open YourAIVoice, or read about downloading an MP3.