Audio in WhatsApp service: what the assistant does
Published on October 06, 2026
When a customer sends a voice note instead of typing, the assistant transcribes it and works from the text.
Replying by voice is a separate step that comes later, in the last stage of setup.
What happens when a customer sends a voice note
The customer taps the microphone and talks.
An audio file lands in the chat, just like a photo or a screenshot.
In the WhatsApp Cloud API, audio arrives as a media message, which the system downloads to transcribe.
Before replying, the assistant needs the content as text, because text is what it reads.
So the audio is transcribed, and the text goes into the chat.
From there, it works just like a typed question.
It books, confirms, reminds, keeps a record or hands over to a person when needed.
What the assistant does with the audio
The audio becomes text, and the assistant reads that text to reply.
The reply is the same as for a typed question: the service, the price, the open slot, the confirmation.
Only the start was spoken.
The audio is also added to the chat record.
Photos, screenshots and audio that come in are saved as files with the right name and folder, and audio is turned into text.
When a voice reply makes sense
A voice reply makes sense when the customer also sent audio.
It also works when the answer is an explanation that is easier to hear than to read.
Text is better when the person will need to look at it again.
Address, price, time and what was agreed are things customers come back to, and in text they find them fast.
A voice reply doesn’t replace the record.
What was agreed stays saved in the chat, as text, for you to check later.
What depends on audio quality
- Noisy audio, recorded on the street or with lots of people talking, produces a patchy transcript. A wrong sentence leads to a wrong answer.
- Long audio is tiring and mixes topics. Splitting it into shorter parts helps.
- Audio with more than one topic, like a booking request and a price question, can make the assistant answer what it understood and miss a point.
- When the transcript makes no sense, the right move is to ask the person to type or hand over to someone on the team, not to guess.
Which stage of setup this belongs to
Audio comes in the next phase of setup.
The service starts with text, along with booking, reminders and records.
Part four of the delivery covers voice and fine-tuning: transcribed audio, voice replies and whatever real use shows is needed.
In short: text comes first, and voice comes later, in part four of the delivery, within the agreed scope.
What you need to decide first
- Whether you want voice replies, and when: for every customer or only for those who sent audio.
- What to do when the audio can’t be understood: ask them to type, bring in a person, or both.
- Who follows up on voice notes that were only partly understood, if any come up.
- Whether audio is in scope from the start or saved for a later stage.
These decisions shape the setup, so they are worth discussing before you sign up.
Our WhatsApp runs the assistant we set up, already live.