◉

Voxa

Text to speech. Nine languages, real voices. Generated on a server, so it sounds the same on every device. Turn a PDF into an audiobook →

0 / 4000 characters

About the voices. Studio voices run Kokoro-82M (Apache 2.0) on our server: one set of weights serves all nine languages, and repeated phrases are cached and come back instantly. On-device voices run Phonon 31m by Gradium (Apache 2.0) in your browser, so they cost nothing and work offline once loaded. Voice names are Voxa's own. Credits