Voice & files
An AI agent that answers WhatsApp voice notes
Many of your customers don’t type, they record. A Gental agent listens to the voice note, turns it into text and answers in writing, as if the customer had typed. Send it a photo or a file and it reads that too.
Updated
How voice notes become text and get an answer
The customer sends a voice note
On WhatsApp (official or linked number), in groups, or with the mic button in your website chat.
Gental transcribes it
The recording becomes text with an engine tuned for Gulf Arabic, helped by your product names and the last few messages of the chat.
The agent answers in text
The text goes through like any question: it searches your knowledge and replies. If an unclear word changes the answer, it checks with the customer first.
Your team sees everything
The transcript shows in the team inbox, and you can play the recording when you need to.
Example: a voice note in Saudi Arabic
The customer recorded a voice note. This is the transcript the agent received:
“Hi, I want to ask about Thursday. Do you have a table for four at nine in the evening?”
Hi! We have family seating on Thursday, and a table for four at nine is open on the bookings calendar. Shall I send you the booking link?
AI that reads images and files too
Customers can send a photo of an invoice, a screenshot of an error or a contract, on every channel.
Images, PDF, Word, Excel
Reads photos (including iPhone HEIC), PDF, Word, Excel, CSV and text files, and answers the customer’s question about them.
Arabic, even handwritten
Keeps Arabic as written and tries to read handwriting. When an image is unclear, it says so.
Masks sensitive numbers
National ID, iqama, passport, IBAN and card numbers are masked in what the agent keeps about the file.
Never added to your knowledge
Customer files are read for that conversation only. They never join your knowledge library or reach another customer.
Where voice notes and files work
| Channel | Voice notes | Images and files |
|---|---|---|
| Website chat | Yes | Yes |
| Official WhatsApp | Yes | Yes |
| Linked number (1:1 chats) | Yes | Yes |
| WhatsApp groups | Yes, switch per group | Yes, switch per group |
| API | No | Yes |
Choose per agent which channels accept voice notes, and turn file reading off for the whole workspace if you prefer.
How long recordings and files are kept
You choose in the privacy settings: delete voice recordings right after transcription, after 30 days, or after 90 days. The same choice applies to customer files: after reading, 30 days or 90 days.
What the agent understood stays in the conversation history so your team can look back. Data is stored in Saudi Arabia, in me-central2, Dammam.
FAQ
Common questions
Does it understand voice notes in Saudi dialect?
Yes. Transcription is tuned for Gulf Arabic and Arabic mixed with English, and the agent replies in the dialect you set for its persona.
Does the agent reply with voice?
No, it always replies in text, which is clearer for the customer and easy to look back at.
How long can a voice note be?
Up to 3 minutes. If it is longer or has no clear sound, the agent asks the customer to send a shorter one or type the question.
Can it read a photo of an invoice or a PDF?
Yes. It reads images, PDF, Word and Excel files on every channel, and masks sensitive numbers such as national ID and IBAN.
Can my team hear the voice note?
Yes. The transcript shows in the team inbox, and a teammate can play the recording until it is deleted under your retention setting.
Keep reading
- WhatsApp AI agentAutomated WhatsApp replies from your knowledge, with handover to your team.
- WhatsApp Linked NumberLink a regular WhatsApp number by QR to answer chats and catch requests in groups.
- Arabic AI chatbotAn Arabic chatbot that understands dialects and Arabizi, and replies in the dialect you choose.
- Human handoff and inboxThe agent hands over to your team, and your team replies from a shared inbox on the same channel.
- Data privacyWhere your data is stored, what reaches the model provider, and how we protect it.
Meet your gentle agent
Build your first agent on your own documents today. Free to start.