Attach PDFs, images, video and audio to a free AI chat

LE PEAK, KTwo’s free multi-model chat, takes files. Drop in a PDF, a photo, a screenshot, a video, a spreadsheet exported as CSV, a folder’s worth of code or a voice memo, and ask about it. Most of it works with the free models, with no account. Here is what each kind of file becomes, which models can read it, and what happens to it on the way.

How to attach a file

Press the attach button in the message box, paste from the clipboard, or drag files onto the chat. Each file appears as a chip above the message; you can attach up to 10 files to one message.

Before you send, the chip tells you whether the model you picked can read that kind of file. If it can’t, say a video for a text-only model, the chip says so and why, and you can switch models: pending attachments follow you to the new model.

What each kind of file becomes

Language models don’t open files. Everything has to arrive as something the model reads: text, images or, for a few models, audio. LE PEAK does that conversion in your browser before anything is sent.

A few details matter in practice:

• Scanned PDFs. A PDF made from a scanner or a phone camera has no text layer, only pictures of pages. LE PEAK notices and sends those pages as images to a model that can see them. For a PDF that does have text, the chip’s Pages button sends its pages as images too, which helps with charts, tables and forms whose layout carries meaning.
• Videos are sampled. A model sees frames, not motion. Sixteen frames are plenty for “what happens in this clip?” or “read the slide at 0:42”, and not enough to count fast events or follow a long film.
• Text files go in whole. Logs, CSV, JSON, Markdown and source code in dozens of languages are read as text, so every model can read them, including the ones that can’t see images.

Which models read what

The model selector marks each model with what it reads. Some models can handle more on their own: Nemotron 3.5 Super VL, Gemma 4, MiniMax and MiMo also read video natively, and GPT Luna and Claude Haiku read PDFs. LE PEAK still sends them frames and text plus page images, the same as everyone else, so a file behaves the same whichever model you pick and costs no hidden per-page fees. The line-up changes as providers add and retire models; the chat’s menu is always current.

Why some models see and others don’t, and why the best of them were built to, is the subject of Natively multimodal models, explained.

Limits

Follow-up questions

Files stay part of the conversation. They go again with every later message in the thread, so you can ask a second and third question about the same PDF without attaching it again, or switch to another model and ask it the same thing.

A model can only read so much at once, its context window. When a long thread with large files no longer fits, LE PEAK leaves the oldest files out, names them in the request instead, and the reply tells you so. What a context window is, and why it fills faster than you might think, is explained in What is a context window?

Files also count toward a model’s daily allowance on the models that have one. LE PEAK reserves room for them before sending (about one token per 750 pixels of image, and 32 tokens a second of audio), then counts what the provider reports. The free models have no daily cap. How the allowances work is in the LE PEAK guide.

What happens to your files

• Your browser does the preparing. Resizing images, cutting frames from a video, reading a PDF’s text and converting audio all happen on your device. Nothing is uploaded anywhere first.
• Re-encoding strips metadata. A re-encoded photo loses its location and camera data before it leaves your browser.
• Only the result is sent, with your message, to the provider of the model you picked, under that provider’s terms. Audio goes only to the MiMo models, the ones that listen.
• KTwo doesn’t store your files, or your messages.
• Your chat history keeps only the files’ names. Chats are saved in your browser; attachments aren’t, so a reopened chat shows which files were there but not their contents.

The full account is on the privacy page. As with any chat, don’t attach anything you wouldn’t want a third-party provider to read: contracts with personal data, passwords, medical records.

Frequently asked questions

Can I chat with a PDF for free?

Yes. Every model in LE PEAK reads a PDF’s text, and twelve models are free with no account. For scanned PDFs, pick one of the free models that can see images, such as DeepSeek V4.1 Flash or Gemma 4 26B A4B.

Which free AI model can read images?

Seven of LE PEAK’s free models read images: Step 5 Preview, DeepSeek V4.1 Flash, GLM 5.3 Flash, Muse Glimmer 30B, Nemotron 3.5 Super VL, Gemma 4 26B A4B and MiniMax M3.1 Flash. None of them needs an account.

Can an AI model watch a video?

Not frame by frame like a person. LE PEAK sends up to sixteen time-stamped frames from a video, which is enough to describe what happens, read on-screen text or find a moment. The MiMo models also hear its soundtrack.

Can I ask questions about an audio recording?

Yes, with a free account, on the MiMo models. LE PEAK sends the first minute of each clip, up to two clips and two minutes per message.

Does KTwo keep my files?

No. Your browser prepares them and sends the result to the model’s provider to produce the reply. KTwo doesn’t store them, and your local chat history keeps only their names.