Reply Wizard
Voice note to text
Transcribe WhatsApp voice messages to text
Reply Wizard is a Chrome extension that turns an incoming WhatsApp voice message into text you can read, search and copy, inside WhatsApp Web on your computer. Click Read transcription on the voice note and the words appear under it in a few seconds. It detects the language itself, handles code-mixed speech such as Hinglish and Tanglish, and can run entirely on your own computer with Whisper, free and with no upload.
How to transcribe a WhatsApp voice message on your computer
To read a WhatsApp voice message as text: install Reply Wizard in Chrome, open the chat at web.whatsapp.com, and click Read transcription on the voice note. The transcript appears under the message in a few seconds, and the voice note is never played aloud.
-
Install the extension
Add Reply Wizard from the Chrome Web Store and open web.whatsapp.com. No account or card is needed.
-
Choose how to transcribe
In the extension settings, under Voice Transcription, choose On-device Whisper (free, nothing uploaded), a cloud engine with your own key, or your Reply Wizard plan.
-
Click Read transcription
Open any chat and click Read transcription on an incoming voice note. The text appears beneath it, with buttons to copy it, put it in your message box, redo it, or change the settings for that one message.
Three ways to transcribe, compared
Reply Wizard can transcribe a voice note in three places: on your own computer with Whisper, through a speech provider you hold a key for, or through a Reply Wizard plan. On-device Whisper costs nothing and sends nothing anywhere; the cloud engines are faster on a modest computer and more accurate on hard audio.
| On your computer (Whisper) | Your own key | A Reply Wizard plan | |
|---|---|---|---|
| What it costs | Nothing, and no limit | Your provider's own rates (Groq, Gemini, ElevenLabs, OpenRouter and more) | Part of the Pro or Power allowance: about 7,000 tokens for a 30-second note |
| Where the audio goes | Nowhere. It never leaves the browser | Straight to that provider, never through us | Through our gateway to Google's Gemini, and it is not kept |
| Setup | Download a model once, from 75 MB (Tiny) to 1.5 GB (Large-v3-turbo) | Paste the key once | Subscribe. There is nothing to configure |
| Works offline | Yes, once the model is downloaded | No | No |
| Best for | Privacy, and anyone with no key | Anyone who already pays for Groq, Gemini, ElevenLabs, OpenRouter or OpenAI | Anyone who does not want to manage keys |
Whisper is the open speech-recognition model published by OpenAI; Reply Wizard runs it in your browser through ONNX Runtime and WebGPU. Source: openai/whisper.
Hinglish, Tanglish and other code-mixed voice notes
Much of WhatsApp speech mixes two languages in one sentence, such as Tamil or Hindi with English words dropped in. Reply Wizard sends every transcription a spelling style, and the default, As spoken, keeps English words in English letters and the rest in its own script, so "office" stays "office" rather than being spelled in Tamil letters.
| Clean audio | With background noise | |
|---|---|---|
gemini-3.1-flash-lite, spelling sample only | 53% | 60% |
gemini-2.5-flash, spelling instruction and reasoning | 90% | 94% |
Measured by us over 162 requests on twelve casual clips in Tamil, Hindi, Telugu, Malayalam, Kannada, Bengali and Marathi, each clean and with noise added, plus real voice notes. The first setup kept no English word in English letters at all in Tamil, Malayalam, Kannada and Bengali; the second fixed all four. The Reply Wizard plan moves to the second setup from version 3.5.7.
Three spelling styles are offered, under Voice Transcription in the settings: As spoken (English words in English, the rest in its own script), One script only (everything in the spoken language's own script) and Latin letters (everything romanised, the way people type in a chat).
What you can do with a transcript
A Reply Wizard transcript is a working piece of text, not just a caption. You can copy it, put it straight into your message box, summarise it, or let a suggested reply answer it, because Reply Wizard reads transcribed voice notes as part of the conversation.
- Summarise a long voice note. Choose a style for the transcript: Normal (up to 160 words), Short (80), Concise (40), or bullet points.
- Reply to it. Suggest Reply reads a transcribed voice note like any other message, so the draft answers what was said.
- Dictate your answer. Hold the microphone in the reply bar and speak; your words land in the message box for you to send.
Where the audio goes
With on-device Whisper, a WhatsApp voice note is transcribed inside your browser and the audio goes nowhere. With your own key it goes straight to that provider. On a Reply Wizard plan it passes through our gateway to Google's Gemini and is not kept: nothing spoken is logged, not the audio and not the transcript.
Reply Wizard transcribes a voice note only when you click Read transcription on it. It never transcribes in the background, and a chat you have not opened is a chat it has not read. The full detail is in the privacy policy.
Questions
Questions about WhatsApp voice to text
Can I transcribe WhatsApp voice messages on my computer?
Yes. Reply Wizard adds a Read transcription control to every incoming voice note in WhatsApp Web, at web.whatsapp.com, in Google Chrome and other Chromium browsers. Click it and the words appear under the voice note.
Is WhatsApp voice to text free?
Yes. On-device Whisper transcription is free and unlimited, needs no account and no API key, and works offline once the model is downloaded. With your own Groq, Gemini, ElevenLabs, OpenRouter or OpenAI key, Reply Wizard is also free and your provider bills you directly.
Which languages can it transcribe?
Reply Wizard detects the spoken language automatically, and you can fix one in the settings if you prefer. Whisper and Gemini are both multilingual, and Reply Wizard has been measured on Tamil, Hindi, Telugu, Malayalam, Kannada, Bengali and Marathi, including voice notes that mix them with English.
Does it work on the WhatsApp phone app?
No. Reply Wizard is a Chrome extension, so it works in WhatsApp Web on a computer, not in the WhatsApp app on iPhone or Android and not in the WhatsApp desktop app.
Does it work in group chats?
Yes. Any incoming voice note in the chat you have open can be transcribed, in a one-to-one chat or a group.
Can it summarise a long voice note?
Yes. Set a summary style for transcripts, from Normal (up to 160 words) to Concise bullet points, and a five-minute voice note comes back as the points that matter.
Is the audio stored anywhere?
Not by us. On-device Whisper keeps it in your browser. With your own key it goes to that provider under their terms. On a Reply Wizard plan it passes through our gateway to Google's Gemini and is not kept, and nothing spoken is logged.
Try it on the next message that arrives
The free version needs no account, no email address and no card, so trying it is cheaper than deciding.