Read WhatsApp Web voice notes as text.
Don't listen to them.

Not as of July 2026. WhatsApp's built in transcripts are a phone feature, turned on in Settings on iOS or Android. Open a voice note on web.whatsapp.com and there is no transcribe option in the message menu. That is the gap this extension fills: it adds a Transcribe button under every voice note in the browser.

No. The extension patches the browser's media play method and returns immediately without ever starting playback, so the audio pipeline is never engaged. This is not muted playback, it is no playback. You can transcribe a voice note in a meeting or on a quiet train without headphones.

Possibly, and we would rather say so than pretend otherwise. To get the audio, the extension has to click WhatsApp's own play control so the file downloads and decrypts. WhatsApp's internal state machine may still mark the note as played and send a played receipt, which is the blue mic on the sender's side. What you reliably get is silence in your room, not invisibility. Suppressing the receipt would mean tampering with WhatsApp's WebSocket frames, which this extension does not do.

The extension is free and open source under the MIT licence. It has no accounts, no subscription and no paid tier. You supply your own ElevenLabs API key and ElevenLabs bills you directly, per minute of audio, at your account's rate. Transcripts are cached per message, so re-reading an old note costs nothing.

From your browser to api.elevenlabs.io, and nowhere else. There is no backend, no analytics and no account, so nothing reaches the developer. Your API key sits in your browser profile's local storage and is only ever sent in the request to ElevenLabs. The romanisation step is rule based and runs entirely on your machine, with no second API call.

ElevenLabs Scribe v2 auto detects and transcribes 90+ languages, including notes that switch language mid sentence. The extension then romanises 25 writing systems locally, covering Devanagari, Bengali, Gurmukhi, Gujarati, Odia, Tamil, Telugu, Kannada, Malayalam, Sinhala, Arabic, Cyrillic, Greek, Hebrew, Georgian, Armenian, Thaana, Hangul, Kana, Han, Thai, Lao, Khmer, Myanmar and Ethiopic. Languages already written in Latin pass through untouched.

Yes, and those need no conversion step at all. Scribe v2 auto detects the language and returns the transcript directly, so a Spanish or German voice note simply becomes readable text. The romanisation layer only comes into play when the language is written in a script other than the Latin alphabet.

No. It is a Chrome extension and it runs only on web.whatsapp.com in a desktop browser. On a phone, use WhatsApp's own transcripts if your language is supported.

Usually one of four reasons: the transcript language set on your phone does not match the language actually spoken, the note is longer than the device will process, on device transcription has not finished downloading its language pack, or you are on WhatsApp Web or desktop where the feature does not exist at all. The first three are settings problems on your phone. The fourth is not fixable in WhatsApp, which is what this extension is for.

Yes, on WhatsApp Web. WhatsApp's native transcripts cover a short list of languages, and Urdu, Bengali, Tamil, Telugu, Punjabi, Marathi and Gujarati are not on it. This extension does not use WhatsApp's transcription at all, so its language coverage is independent of what your phone offers.

That is where native transcription tends to fall over, because it makes you pick one transcript language at a time. Scribe v2 auto detects and handles mid sentence switching, and the extension romanises only the non Latin runs, so ok भेज दो, спасибо, 谢谢 reads as ok bhej do, spasibo, xie xie.

Yes. Every transcript has a toggle that switches between the romanised version and the original script, and you can set the default to show either one or both stacked. Copy takes whichever version is on screen.

No, and the difference matters. Translation changes the meaning into another language. This changes the script only, so the words stay Hindi and just become readable in Latin letters. If you cannot understand spoken Hindi, a Hinglish transcript will not help you. If you understand Hindi fine but cannot read Devanagari quickly, it is exactly what you need.

Yes. That is the main reason it exists. A Hindi voice note comes back as Hinglish, so सर वो ऑर्डर अभी तक डिलीवरी नहीं हुआ reads as sir wo order abhi tak delivery nahi hua. A toggle flips the same transcript back to Devanagari whenever you want the original script.

Honestly, weaker than the Indic ones. Arabic script does not write short vowels, so a word like کل could be kal, kul or kil, and the extension leans on a dictionary of common words to get most of it right. Urdu comes out well, Arabic is decent, Persian is romanised with Urdu vowel colouring, and Pashto is poor. Devanagari, Bengali, Gurmukhi, Gujarati, Odia, Tamil, Telugu, Kannada, Malayalam and Sinhala are the strongest modules.