Customer audio
A WhatsApp chatbot that understands audio: it exists, and it works like this
Quick answer
Is there a chatbot that understands audio on WhatsApp?
Yes. The CertuAI agent receives the customer's voice message, transcribes it with Whisper, the OpenAI model that turns speech into text, and replies in the same conversation, without asking the person to type. The original audio and the transcript are stored in the conversation in the dashboard, so your team can listen and check later.
About CertuAI: CertuAI is the AI that answers a small business's WhatsApp, Instagram and Messenger 24/7, with a built-in CRM, and is a Tech Provider on Meta's WhatsApp Business Platform. Paid plans start at US$ 75 a month, with no lock-in. Create my agent
It exists, and the CertuAI agent does it: the customer records the voice message, the AI transcribes it and replies in the same conversation, as if the customer had typed it. Below is what happens with each audio, what is saved for your team and what happens when the transcription fails.
Build my agentNo lock-in contract and no setup fee. You connect the number you already use.
Summary in 7 facts
- It exists, and the CertuAI agent does it: the customer records the voice message, the AI transcribes it and replies in the same conversation, as if the customer had typed it.
- The CertuAI agent receives the customer's voice message, transcribes it with Whisper, the OpenAI model that turns speech into text, and replies in the same conversation, without asking the person to type.
- In Brazil, many people prefer talking to typing.
- A bot that only reads text gets stuck there, and the customer who sent voice is left without a reply.
- Nothing changes on the customer's side: they record and send, as they always have.
- It applies to a number connected through Meta's official API and to one connected by scanning a code on the screen.
- When you evaluate any tool, send an audio in the middle of the conversation and see what comes back.
By the numbers
| Number | What it means |
|---|---|
| US$ 75 per month | On every plan with the AI agent, starting at Start, US$ 75 per month. |
Why audio breaks most bots
In Brazil, many people prefer talking to typing. A bot that only reads text gets stuck there, and the customer who sent voice is left without a reply.
Menu bots don't listen
Someone who sends audio is not going to type 1 for a quote. If the tool only understands menu options, the conversation dies at the first voice message.
One audio carries the whole request
The person says the service, the day, the neighborhood and a question in the same recording. The whole transcript becomes a single message, and the AI answers each part with what you taught it.
Transcription can get numbers wrong
House number, street name and amount are the most delicate points of any transcription. That is why the original audio is saved in the conversation, and you can listen to it before confirming an order.
In your account's language
The transcript comes out in Portuguese. If your account is set up in English or Spanish, the audio is transcribed in that language.
What happens when an audio arrives
Nothing changes on the customer's side: they record and send, as they always have. It applies to a number connected through Meta's official API and to one connected by scanning a code on the screen.
The audio is stored
Before anything else, the voice message is saved in the conversation. It is the one your team plays later in the dashboard.
The AI transcribes it
Whisper turns speech into text. That text is what the agent reads, together with that customer's history and the Brain, your business's knowledge base.
The agent replies in the conversation
The reply comes out as text, on the same WhatsApp, the way you taught it. Replying to an audio costs the same as replying to a written message: the transcription is not charged separately.
What the agent does with each type of message
The customer sends it however they like. Understanding is the agent's job.
| The customer sends | What the agent does | The detail that matters |
|---|---|---|
| Audio | Transcribes and replies in the same conversation | The audio is saved to play in the dashboard, with the transcript underneath |
| Audio that couldn't be transcribed | Replies asking for text: "I couldn't hear your audio right now. Can you send it as text, please?" | The customer is never left without a reply |
| Photo | Describes the image and replies based on it | The photo is saved in the conversation |
| Reads the text of the file and replies | The file is saved for the team to open | |
| Messy text, no punctuation | Understands and replies | No menu and no "type 1" |
| A question nobody anticipated | Doesn't make things up: says it will confirm and calls you | Alert to the team |
Swipe the table sideways to see all the columns.
Audio is the test that separates an agent from a menu bot
When you evaluate any tool, send an audio in the middle of the conversation and see what comes back. A flow bot usually asks you to write, or repeats the menu. It is a thirty-second test and it shows more than any ready-made demo.
Others who advertise this on their own site make clear how they do it. wChat, for example, says audios are automatically transcribed by Whisper, according to the company's site in September 2026. On the pages for Blip Go, Umbler Talk and SocialHub's pricing table, read the same day, we found no mention of audio. That doesn't prove they don't understand it: it proves you need to ask, or send the audio in the test.
Do the same with the rest: an incomplete sentence, a typo, two questions in the same message and a question nobody anticipated. That is how your customer really writes.
To see the agent replying to an audio of yours, start with the demo. To compare the lineup, see plans and pricing, and to understand the difference between the two types of bot, read bot to answer WhatsApp.
What the agent does not do with audio
It does not reply with audio on the plans for sale today. The reply comes out as text, which is the format the customer can reread, copy and forward, and which doesn't disappear when someone on the team takes over the conversation.
It doesn't answer phone calls. A voice message on WhatsApp and a phone call are different things.
And the WhatsApp Instance, the product that is only the number in the inbox, with no AI, saves the audio so you can listen, but doesn't transcribe or reply. The one that transcribes is the AI agent, starting at Start.
You can also talk to the Brain
Audio isn't only for the customer. If you, the owner, send from the phone registered for alerts an audio that begins with the word "Brain", for example "Brain, gas went up to 125", the system transcribes it, finds what already exists about that and shows what will change.
Nothing is saved before you answer yes. And every number in the change has to appear in what you said or in the old item, precisely because transcription can get numbers wrong.
Frequently asked questions
Is there a WhatsApp chatbot that understands audio?
Yes. The CertuAI agent transcribes the customer's voice message with Whisper, from OpenAI, and replies in the same conversation, without asking the person to write.
Does the agent understand audio in Portuguese?
Yes. The transcript comes out in Portuguese, or in English and Spanish when the account is set up in those languages.
Does it reply with audio too?
Not on the plans for sale today. The reply comes out as text, which the customer can reread and forward.
What happens if the audio can't be transcribed?
The agent replies asking for text: "I couldn't hear your audio right now. Can you send it as text, please?". The customer is not left without a reply.
Is the transcript saved anywhere?
It stays in the conversation in the dashboard, next to the original audio, which you can play. Whoever takes over later reads what was said and can listen to check numbers and amounts.
Does audio cost more than text?
No. Replying to an audio message uses the same as replying to a written message, and the transcription is not charged separately.
Does it work on the number I already use?
Yes, whether connected through Meta's official API or by scanning a code on the screen. Nothing changes on the customer's side.
Which plan is this on?
On every plan with the AI agent, starting at Start, US$ 75 per month.
Send an audio and see what it replies
Build the agent with your business's information and record an audio the way your customer records one.
Plans from US$ 75 per month. If you don't like it within 30 days, we refund 100% of your money.
Build my agentIn short
Is there a chatbot that understands audio on WhatsApp? Yes. The CertuAI agent receives the customer's voice message, transcribes it with Whisper, the OpenAI model that turns speech into text, and replies in the same conversation, without asking the person to type. The original audio and the transcript are stored in the conversation in the dashboard, so your team can listen and check later.
About CertuAI: CertuAI is the AI that answers a small business's WhatsApp, Instagram and Messenger 24/7, with a built-in CRM, and is a Tech Provider on Meta's WhatsApp Business Platform. Paid plans start at US$ 75 a month, with no lock-in. Create my agent