Talk, see, and listen

You've been typing. But you can also show an AI a photo, talk to it out loud, or share your screen — and it takes all of that in. How does a text machine deal with a picture?

Many senses, one window

It's the same idea you've already seen, stretched wider. A multimodal AI can accept more than typed words — a photo of your fridge, a screenshot of an error, a spoken question, a PDF. It doesn't matter which: each one gets turned into tokens the model can read, and lands on the same shared window alongside your text. Snap a photo of a plant and ask what's wrong with it; paste a screenshot and ask what the error means; speak instead of type on a walk. Different senses in — one conversation.

Show, don't spell out

This changes how you ask. Instead of describing a problem in careful words, just show it: the photo, the screenshot, the receipt. It's usually faster and less error-prone than typing it all out — a picture really is worth the paragraph you didn't have to write. Voice, likewise, is often the easiest way in when your hands are busy or the question is quick.

The gist

• —Multimodal AIs take in images, voice, screenshots, and documents — not just typed text
• —Each input becomes tokens and joins the same context window, so you can mix them freely
• —When something's easier to show than to describe, show it — it's faster and clearer