> For the complete documentation index, see [llms.txt](https://docs.artific.nl/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.artific.nl/en/for-everyone/assistants/files-and-voice.md).

# Files, images and voice

Attaching documents to a question, and speaking instead of typing.

## Attaching a file

Where the assistant allows it, a paperclip button sits beside the message box. You can also drag files straight onto the conversation.

Hovering the paperclip tells you what this assistant accepts: which formats, the maximum size per file, and how many files you may attach.

{% hint style="info" %}
Scanned PDFs are supported. The platform reads the text off the scanned pages for you, so you can attach a scan and ask about it like any other document. Recognition works best on a clean, straight scan; a photographed or skewed page gives poorer results.
{% endhint %}

An attached file belongs to that conversation. It is not added to your organisation's knowledge base, and other people do not see it.

### What to ask once a file is attached

* *"Summarise this in five bullet points."*
* *"What does this say about cancellation terms?"*
* *"Pull the dates and amounts into a table."*
* *"Rewrite section 3 in plain language."*
* *"Compare these two documents and list the differences."*

Attaching several files at once and asking for a comparison is one of the most useful things you can do here, and one of the least obvious.

### If you need it every time

A file you attach is used once. If the assistant should *always* know something, that belongs in your organisation's knowledge base. Ask your administrator. If several conversations need the same files, use a [project](/en/for-everyone/projects.md).

## Images

Where an assistant can read images, you can attach a photo or a screenshot and ask about it: what an error message says, what a diagram shows, what is on a form.

## Speaking instead of typing

A microphone button dictates your message. Your browser asks permission the first time. If you decline, you will need to allow it again in your browser settings.

Speaking is quicker than typing for anything long, and unlike typing it does not discourage you from giving the assistant proper context.

## Voice conversations

Some assistants support a spoken conversation, where you talk and the assistant talks back.

A few things make it work better:

* Check your microphone before you start.
* Speak clearly and at a normal pace.
* Wait until the assistant has finished before you reply. Talking over it does not work.
* Expect shorter answers than you would get in writing. That is deliberate: long spoken answers are hard to follow.

If the assistant is embedded in a page on another website, that page also has to allow microphone access. When it does not, the assistant tells you.