Media
2 items
Description
Gives any page a voice. Local TTS on your GPU — text never leaves your device. Page Voice gives any web page a voice — with a neural text-to-speech engine that runs entirely inside your browser, on your own GPU. READ ANYTHING Select text on any page, right-click, and choose "Read aloud" (or press Alt+R). A natural neural voice starts speaking in about a second. AUTO-READ AI CHATS On ChatGPT, Claude, and Gemini, flip on Auto-read in the small status pill: streaming replies are spoken sentence by sentence, while the text is still being generated. Synthesis runs faster than playback, so long answers play without gaps. 100% ON-DEVICE There is no server and no API key. The TTS model (Matcha-TTS + HiFi-GAN, ~90 MB) is downloaded once from Hugging Face, cached, and then runs locally via WebGPU and multithreaded WebAssembly (LiteRT.js). The text you read is never sent anywhere — it cannot be, there is nothing to send it to. - Works offline after the one-time model download - Free, no account, no tracking, no analytics - Latency badge shows the real-time factor and compute backend - Open source: https://github.com/john-rocky/page-voice