Support the service
Vox2Text is free and ad-free. If you find it useful, you can help keep it running with a voluntary donation.
Frequently asked questions
How does it work?
The site runs Whisper, a speech recognition model, right in your browser. The first time it downloads the model (75 to 550 MB depending on quality); after that it stays cached and starts instantly.
Are my recordings safe?
Yes. Files are never uploaded to any server: they are read and transcribed on your device. The site uses no cookies and no tracking.
Where do I find WhatsApp voice messages?
WhatsApp Web or Desktop: hover over the voice message, open the menu and choose “Download”.
Android: folder Android/media/com.whatsapp/WhatsApp/Media/WhatsApp Voice Notes.
iPhone: long-press the voice message › Forward › Share › Save to Files.
Is the language detected automatically?
Yes: with “Detect automatically” the site listens to the first seconds and picks the language. If it gets it wrong, change it under the text and press “Transcribe again”, or choose it beforehand from the menu.
Which quality should I pick?
“Balanced” works almost always. “Fast” downloads less but makes more mistakes. “Best” is the most accurate and needs a browser with WebGPU (up-to-date Chrome or Edge).
Which browsers are supported?
Up-to-date Chrome, Edge and Firefox on computers or Android. On iPhone you need iOS 18 or later. On computers with a graphics card transcription is much faster.
Can I transcribe long recordings?
Yes: up to 200 MB and 3 hours per file on a computer, 1 hour on a phone. Long files take longer: keep the page open (the screen stays on automatically) and follow the time-left estimate. The text appears as it goes and you can copy it at any time.