Voice and images
How to translate WhatsApp voice messages
Translate a WhatsApp voice message by transcribing the audio first, then translating the text. Chrome, iPhone, and Android steps, plus format limits.
A voice message is audio, not text, so nothing can translate it directly. The practical path has two steps: turn the recording into text, then translate that text into the language you read. Skipping the first step means guessing at sounds, which is exactly what a translation should not do.
You can do this on WhatsApp Web in Chrome, on iPhone with Shortcuts, or on Android with the native app. Each path handles one message that you choose. Your reply is still written and sent by you inside WhatsApp.
Transcription comes before translation
Speech recognition listens to the audio and writes down what it hears. The translation model then works on that transcript, not on the sound. If the transcript is wrong, the translation inherits the error, so the transcript is the part worth checking.
Read the transcript first and ask whether it matches what you expect from the conversation. When the original and the translation disagree, go back to the audio before you trust either one, and never forward a translated voice message as if it were a quotation.
What changes between devices
On the web, the result sits in a card below the message you selected. On a phone, you share one audio item to the shortcut or the native app and read the result on its own screen.
No phone version injects a menu into WhatsApp. The share sheet is the entry point, and each share carries one item. The platform steps at the end of this page show the exact path for each device.
Formats, waiting time, and accuracy
Accepted audio formats are OGG/Opus, MP3, WAV, and M4A. Other formats are rejected instead of being guessed at, so if a recording will not load, check the container before trying again. Voice tasks run asynchronously: the audio is submitted, then the result is polled, so a long recording does not block the screen.
Recognition can mishear names, accents, background noise, and strings such as model numbers or prices. Treat every number in a transcript as unverified until you replay the audio or confirm with the sender. No tool can promise perfect accuracy, and this one does not.
What to check before you rely on the result
Read the transcript against the audio before the translation. Names, quantities, prices, dates, and model codes are the first things to check; a fluent sentence can still contain a wrong number.
If the audio is unclear, ask the sender to repeat the number in text. For a quotation or a payment instruction, confirmation is part of the task, not an optional extra.
When not to use AI translation
Do not use a machine transcript as a certified record, a contract, or a legal interpretation. It is a reading aid.
If the recording is inaudible, the file format is unsupported, or the message is mostly music or noise, stop and ask for text instead of guessing.
Do it on your device
Web
- Open the voice message in WhatsApp Web on desktop Chrome 120 or later.
- Open the message menu and choose identify and translate.
- Read the transcript and translation in the result card below the message, then copy the reply you want.
If the page cannot expose the audio, choose the local file option.
iPhone
- Share the single voice message from WhatsApp to the inWind WA Helper shortcut.
- On first run, pick the interface language and the target language.
- Copy the translation from the result page back into WhatsApp.
Android
- Share the voice message from WhatsApp to the native app.
- Set the interface language and the translation target separately if needed.
- Copy the result back into the chat and send it yourself.
Frequently asked questions
Can I translate a voice message without transcribing it?
Not reliably. Translation works on text, so the audio is transcribed first. The result includes both the recognized original and the translation.
Which audio formats work?
OGG/Opus, MP3, WAV, and M4A. Other formats are rejected.
Why does voice take longer than a text message?
Voice runs asynchronously: the audio is submitted, then the result is polled. Text and image tasks wait up to 8 seconds synchronously; voice continues in the background.
Does it translate every voice message in the chat?
No. It handles the single message you select. It does not scan the chat and does not send replies.
Give your next WhatsApp message to inWind
Understand foreign-language messages from your customers and prepare a reply in the language you know best. Text, images, and voice notes included.