Voice notes, images and documents

An AI chatbot for WhatsApp voice messages, screenshots and PDFs

Customers who talk instead of typing, or send a photo of the problem, get a written reply built on what they sent, on WhatsApp, Instagram, Messenger and web chat. Replies stay in text: no spoken answers and no phone calls.

  • Voice notes transcribed on all four channels
  • Reads screenshots, receipts and scanned PDFs
  • Per-channel switches for voice, images, video and documents
  • Transcript and extracted text under each message in the inbox
Jordan M.WhatsApp
0:14
“Hi, does the Sage planter come in a bigger size? And does it have a drainage hole?”
Yes, the Sage planter comes in 8 and 12 inch. Both have a drainage hole with a cork plug. [Source: Product catalog]
AI · 09:41 AM ✓✓
The voice note is transcribed, then answered in text like any typed message.

WhatsApp voice notes get a written answer

A customer records 40 seconds while walking the dog and asks about two things. Risolva transcribes the note, the AI replies in text to what was asked, and the original stays playable in the inbox. The spoken language is detected automatically, so a note in Spanish gets a Spanish reply.

Screenshots, photos and scans become usable text

Each image is read for every line of text it holds, plus a short summary of the key details. Order confirmations, error screens, shipping labels and serial number plates all come through. Scanned PDFs and Word files are read too, so a customer can send the invoice instead of retyping it.

You decide which media the AI handles

Every channel linked to a bot has its own four switches. You might let the AI answer voice notes on WhatsApp and leave photos of damaged items to your team. Anything switched off still arrives in the inbox. The AI leaves it alone and doesn't mark it as seen.

What happens after the file arrives

The steps are the same on all four channels, whether it's a 0:14 voice note on WhatsApp or a photo of an order confirmation in an Instagram DM.

  1. Saved to the conversation first

    The voice note, photo or file is downloaded and stored with the chat before the AI reads it, so your team can play or open it whatever the AI does next. While it's being processed, the message shows Analyzing... in the inbox.

  2. Turned into text

    Voice notes go through speech-to-text. Images are read for all visible text and summarized. PDFs, Word .docx files and plain-text formats are read directly, and scanned PDFs fall back to the vision model.

  3. Read together with the caption

    If the customer typed something with the file, the AI sees both. A photo of an order confirmation sent with 'I got the wrong size, can I swap it?' is answered with the order details in mind.

  4. Answered, with the text kept on record

    The reply uses the extracted text like any other message, so the order number on a confirmation photo doesn't need typing out again. Under the message, your team can open Show transcript, Show image description or Show extracted text.

Sam T.Instagram
Got the 8 inch but I ordered the 12. Can I swap it?
02:16 PM
Thanks, Sam. I can see order #5132 for the 8 inch Sage planter. Unopened items can be exchanged within 30 days, so you're covered. [Source: Returns and exchange policy]
AI · 02:17 PM ✓✓
Photos, screenshots, PDFs and Word files are read for text, so the reply uses what was sent.

Formats, switches and limits

What each channel accepts, what gets read, and where the controls live.

Voice notes and mediaIn the Risolva app
Channels
WhatsApp, Instagram DMs, Facebook Messenger and web chat. Web chat visitors get an attach button and a voice note button; recording needs an https page and microphone permission.
Voice notes
Transcribed with automatic language detection, then answered in text. No spoken replies and no phone calls.
Images
JPEG, PNG and WebP, including WhatsApp stickers. All readable text is extracted, plus a short summary of key details.
Documents
Read: PDF (text-based or scanned), Word .docx, .txt, .csv, .json, .xml and .md. The web chat attach button takes PDF, Word, Excel, PowerPoint, .txt and .csv files.
Not read
Videos and Instagram story mentions show in the inbox but aren't analyzed. Older .doc files, Excel and PowerPoint files get a reply saying that file type can't be opened.
Media handling switches
Bot > Channels > Media handling, per linked channel: Respond to voice notes, Respond to images (photos and stickers), Respond to videos, Respond to documents. All on by default.
When a switch is off
The file still shows in the inbox and can be played or opened. The AI doesn't transcribe or read it, doesn't reply, and doesn't mark it as seen.
Web chat file size
Max file size (MB) in the web chat widget's Privacy & data settings, 10 MB by default.
The full guide · 4 min readHow to handle customer voice notes, photos and PDFs in chatCustomers send voice notes, screenshots and PDFs instead of typing. How an AI assistant transcribes and reads them, replies in text and leaves video to your team.Read the guide

What teams ask about AI for WhatsApp voice messages

No. Risolva always answers a WhatsApp voice message in text. The customer's voice note is transcribed, the AI replies to the transcript the way it would to a typed question, and the answer arrives as a normal message. It doesn't speak and doesn't take phone calls. The original voice note stays playable in the inbox, with its transcript one click away.

Bring a voice note and a scanned PDF to your demo

Book a demo with a few files your customers really send, such as a long voice note, a screenshot or a scanned invoice. We'll run them through a test bot so you can read the transcript, the extracted text and the reply together.