Text-to-Speech
A private, browser-based voice synthesis tool — runs the full voice model on your device, sends nothing to any server, and produces clean audio from any text you give it.
Features
Fully On-Device
The voice model downloads once to your browser's storage (~92 MB) and runs locally every time after. Nothing is sent to any external server — your text and your audio stay on your machine.
Many Voices
Choose from a full library of American and British English voices — male and female — each with distinct character and quality. Select the one that suits your material.
Pronunciation Control
Force any word to be spoken a specific way using inline IPA phoneme notation. A built-in converter helps you generate the correct phoneme string from plain English spelling.
Stream & Auto-Split
Stream mode plays each sentence as it's generated for immediate playback. Auto-split handles long text by processing it sentence by sentence, preventing cutoff on extended passages.
Export as WAV
Save any generated audio directly to your device as a clean WAV file, named automatically from the first words of your text.
×
Desktop and laptop only
Aīris Speak works in desktop and laptop browsers only — Chrome, Edge, Firefox, or Safari on a Mac or Windows computer. Phones and tablets are not supported, because the voice model requires the processing power and storage of a full computer.
One download, then it is yours
The first time you open Aīris Speak, the browser will download the voice model (Kokoro 82M, around 92 MB). This happens once. After that, everything runs entirely on your machine — no internet connection needed, nothing sent anywhere.
What happens when you open it
The page will show a loading indicator while the model downloads. When it is ready, type any text into the box and press speak. That is all there is to it.