> ## Documentation Index
> Fetch the complete documentation index at: https://docs.min-ai.ar/llms.txt
> Use this file to discover all available pages before exploring further.

# Files, images and audio

> Send it documents, photos and voice notes; your agent reads them, and sends you back files it makes.

Your agent understands what you send it: it reads PDFs and spreadsheets, looks at photos and listens to your voice notes. And it can also send you files it makes itself: a PDF, a spreadsheet, a chart, a screenshot.

## Sending it files

Send it photos, documents (PDFs, spreadsheets, text…), audio or video on your channel, like you would to any contact. In the minai chat, with **Attach**, by pasting an image, or with **Camera** on your phone.

* "Summarize this PDF for me."
* "Turn this spreadsheet into a chart."
* "What does this invoice say? Note down the due date."
* "Here's the contract: is anything odd in the termination clauses?"

### Limits

* Up to **15 MB per file**.
* In the minai chat, up to **10 files per message**.

If a file is bigger, it doesn't reach it. On Telegram, WhatsApp and Slack your agent finds out it was skipped, so it can tell you. In the minai chat the file doesn't upload and you see “Not uploaded: contrato.pdf”.

## Photos and images

* **It reads them:** a license plate, a receipt, a screenshot of an error, a sign, a menu.
* **It sends you images:** screenshots of web pages and charts it draws from your data.
* In the minai chat, tap an image and it opens full size, with arrows to move between several.

<Warning>
  Your agent **doesn't generate images** with AI (drawings, illustrations, made-up photos). The images it sends you are screenshots or charts it makes from your data.
</Warning>

## Files it makes

* "Make me a CSV with these expenses."
* "Create the report as a PDF."
* "Make me a bar chart with this month's sales."
* "Save this page as a PDF for me."

It sends them to you in the same chat.

## Voice notes

Every piece of audio you send it (a voice note or an audio file, on any channel) **is turned into text automatically** before your agent sees it. You talk to it and it understands you.

* 🎤 "Remind me on Monday at 9 that I have the dentist."

Transcription uses your plan's **audio minutes per month**. You see them in **Settings** → **Plan and usage** → **Transcribed audio**. How many minutes each plan includes: see [Plans](/en/reference/plans).

If the month's minutes run out, you get a notice and your agent asks you to type to it until the next month starts. Everything else keeps working.

<Note>
  Your agent **always answers you in writing**: it doesn't reply with voice notes.
</Note>

## Reactions

Reactions work both ways: if you react to one of its messages, it notices (and usually doesn't reply). It can react to yours too, for example with a 👍 as an answer.

<CardGroup cols={2}>
  <Card title="Browse the web" icon="globe" href="/en/reference/agent/browse-the-web">
    Screenshots, PDFs of pages and more.
  </Card>

  <Card title="Plans" icon="credit-card" href="/en/reference/plans">
    Each plan's audio minutes.
  </Card>
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.