Voice notes with context
Customers talk the way that feels natural. Meteor turns the voice note into text, keeps it with the conversation and hands it to the Met before it replies.
Meteor transcribes the WhatsApp voice notes that reach your business number automatically, when the message comes in through an official connection that delivers the file to the system. Meteor stores the voice note, gets its text and adds it to the contact context before the Met runs, so the conversation can continue without the customer turning their audio into a written message. Voice notes are transcribed with Spanish as the expected language.
Two different searches land here. If you want to read the voice notes in your own WhatsApp, the WhatsApp app does that on your phone: Settings, Chats, Voice message transcripts (WhatsApp official blog, checked on October 29, 2026). If you run a business and want each transcript to keep the sale or the case moving, that is what Meteor does.
Transcribing what the customer sends and replying with a voice note are different capabilities. If you also want the Met to speak, compare the voice library and authorized cloning (in Spanish).
Who builds Meteor: the team is certified by OpenAI as an API Deployment Practitioner.
How it works
Through the official WhatsApp connection. If you answer from the WhatsApp Business app on your phone, coexistence keeps the same number on the phone and in Meteor, when Meta enables it for your number.
Once the number is connected, new voice notes are transcribed as they arrive, as long as you have Energy available and within the limits explained below.
Open the conversation: the text appears under each audio. If the note is long, it expands with one click.
With a Met assigned to the channel, the Met uses that text to continue. Without a Met, your team reads the transcript and replies from the same inbox.
Examples, flow and limits
The examples are translated; the audio itself arrives in Spanish. What the Met does next depends on what you have connected and allowed.
«Hi, for tomorrow at one: two set lunches and a lemonade, to Calle 45 with 20.»
The text sits under the audio. With your menu and delivery zones loaded, the Met confirms products, quantities and address before passing the order on.
«My order arrived but one piece is missing. The number is 1234.»
If your store is connected, the Met can look up that order with the number now in writing. If it is not, the conversation can go to a person with the complaint already in text.
«Confirming the price we discussed: it stays at 180,000 and you send me the receipt on Friday.»
The note the agent sends the customer from the inbox is transcribed the same way. Whoever picks up the case reads the agreement in the conversation without listening to the audio.
A flow you can test
The audio, the transcript, the recognized fields and the follow-up action leave separate evidence, so you can review each stretch without mistaking a reply for a real update.
Example to test (spoken in Spanish): «We are six people in sales and want a demo to handle WhatsApp». Meteor keeps the voice note as an attachment on the contact.
The transcript is tied to the audio and enters the context before the Met prepares its reply. If no text comes back, the process does not claim it understood the message.
With that example, the Met can separate intent «demo», team size «6» and channel «WhatsApp», as long as those fields are defined in your CRM.
If it has permission and the required data is there, the Met can update the contact and continue over WhatsApp. If one detail is missing, it can ask for just that; if the case needs judgment, it leaves it ready for a person.
What to check: that the audio is still attached, that the text shows up in the context, that the CRM only shows the authorized fields and that the reply changes when a detail is missing. See the scope of the WhatsApp Mets.
Where it helps most
The transcript is useful when it enters a defined process, with permissions and tools for the Met.
Turns a long spoken explanation into usable context to answer the question or hand it over, without asking the customer to type everything again.
Captures the product, quantity, date or need someone said out loud, so qualification can continue under the rules of your sales process.
Keeps the spoken description in the history so a person or a Met can pick the case up with the same information.
The customer dictates the order in a voice note and the text is ready for the Met to confirm products, quantities and address under your business rules, without asking them to type it.
See how it works in restaurantsThe patient says what they need and when they can come. With that text, the Met offers the open slots in your calendar, or leaves the case to a person if it needs clinical judgment.
See how it works in healthcareIf you also need contacts, stages and owners shared across the team, see the Meteor CRM for WhatsApp (in Spanish).
Both halves of the conversation
It is not only what the customer sends. The voice notes your team replies with also become text, so you can review a whole conversation without listening to every audio.
When an agent records a voice note from the Meteor inbox, that note is transcribed just like the customer’s, and the text sits next to the audio in the same conversation.
If the owner replies with audio from the WhatsApp Business app on their phone, that note is transcribed too. For that, the number must be connected with coexistence: the phone app and Meteor on the same line.
In the inbox each voice note shows its waveform, you change the playback speed with one click and the text appears under the audio. If the note is long, the text expands with one click, without playing it.
Transcribing your team’s notes consumes Energy, just like the customer’s. The Met’s own replies are not transcribed, because their text is already in the conversation.
Clear limits
What it connects
Instagram and Messenger, if you connect them, go through the same transcription as WhatsApp.
A real operation behind this solution.First-person customer stories, with the source and the limits of each case kept visible.
24/7service from first message to closed sale, according to Julián
Julián Chavarro, a marketing agency owner, moved from rigid bots to an AI agent that serves and sells 24/7 with text, audio and images.
Julián Chavarro· Marketing agency owner
Plans From US$99 a month. Every new account starts with a 7-day trial and US$5 of Energy, no card required.
Before you set it up
It is not that tool. Meteor transcribes the audio that reaches a business WhatsApp number connected to Meteor through the official connection, and the voice notes your team records from the Meteor inbox. A chat that is not connected never passes through Meteor, so its audio is not transcribed. For your own chats, the WhatsApp app offers voice message transcripts on the phone, under Settings, Chats, Voice message transcripts.
Turning the voice note into written text. In Meteor the text does not replace the audio: it sits under it, in the same conversation, so you can read it without playing it, search it later and hand it to the Met as context.
Yes, when the audio reaches a business number connected to Meteor. Nobody has to forward or request it: the voice note is transcribed when it arrives, usually before the Met replies, as long as you have Energy available. What happens with audio over 24 MB or when transcription fails is under «What happens when a voice note cannot be transcribed».
A dedicated speech-recognition model does it, separate from the one the Met uses to reply. Meteor requests the text, stores it next to the audio and puts it into the conversation context so the Met can use it when replying.
Voice notes that come in through the channels are transcribed with Spanish as the expected language. If many of your customers speak English or another language, test it with real audio in the demo before relying on it.
WhatsApp voice notes and common audio files such as MP3, M4A, OGG, WAV or WebM. Formats the transcription provider does not accept as they are, like the AAC some phones send, are converted before being transcribed.
Yes. If you connect Instagram or Messenger to Meteor, the voice notes that arrive through those channels go through the same transcription as WhatsApp ones, and the text sits next to the audio.
No. When the transcription finishes correctly, the text enters the contact context and the Met can continue from there. If something was unclear, it can ask for the missing detail instead of making the person repeat everything.
Yes. The extracted text is tied to the audio and is added back whenever the conversation history loads. The original attachment is kept too.
It is built for conversational voice notes. The technical limit of this path is 24 MB; a file larger than that is not transcribed automatically.
Yes, if the Met has the permission and the tools to do it. The transcription provides the text; what happens next depends on the process, the integrations and the limits you configure.
Yes. If you answer from the WhatsApp Business app on your phone today, you can keep using it: with coexistence, the same number is connected to Meteor and your contacts and up to six months of conversations are imported. From the moment you connect the number, new voice notes from customers are transcribed, and the voice notes you reply with from the phone are too. Coexistence depends on Meta enabling your number, and files older than 14 days do not come over from the imported history.
Yes. The Meteor public API returns each transcribed audio with its text in the message attachment, in the attachments[].transcript field, and the message received and message sent webhooks carry the same field: both the customer’s audio and your team’s voice note. The message text still says «[Audio]», so an integration you already have does not change. If transcribing a customer’s audio takes longer than 30 seconds, the webhook goes out without it and the text is available when you query the messages through the API.
In a demo we look at which voice notes you receive, what should be extracted and what the Met can do with that information next.
