AI images and video
Episode 7 · 2 minutes 42 seconds · Free to learn · Words printed on screen · Watch above, then try the exercise below.
What you'll learn
- How AI makes a picture from a description — it builds a new one from patterns; it does not fetch a photo
- How to fix a first try by changing one thing at a time in your words
- What AI pictures are wonderful for — cards, newsletters, invitations — and what they are not: photographs
- The tells that give AI pictures away: extra fingers, melted words, shiny skin, wrong-way shadows
- How short AI video clips work, what they're good for, and the honesty rule for sharing anything AI-made
Key points
Describe a picture in plain words and AI builds it — new, from the patterns it learned from millions of pictures. In the video, one sentence ("a watercolor picture of a sunny community garden, tomato plants in neat rows, morning light") paints a whole garden. First tries are rarely right: our fence leaned and one tomato plant thought it was a tree. So ask again, changing one thing at a time — same garden, late afternoon, golden light, straight fence. AI pictures are made, not taken: nobody stood there and no camera clicked, so learn the tells (extra fingers, melted words, skin a little too shiny, shadows falling the wrong way) and keep lettering out of the picture — words are the weakest part; add your own real words afterward. The same kind of ask makes a few seconds of video: a mood, a greeting, a header — not a whole film. And the honesty rule governs all of it: if AI made it, say so, and never pass a made-up picture off as a real photo of a place, an event, or a person. Camp rule: describe it clearly, check it closely — and say it was made with AI.
Try it
- Describe a picture for something real in your life — a card, a newsletter, an invitation.
- Make two versions, changing one thing between them. Pick the better one.
- If you share it, say it was made with AI.
What does AI see in a picture?
Lesson 7 said AI doesn't see the way you do — it matches patterns. This model labels what's in a picture, entirely in your browser. Try it on our lesson photos: see where it's right, and where it's confidently wrong.
How this works: the model file downloads once from Hugging Face, a public model library, and is then stored on your device. The photos are ours — the demo only ever looks at these four pictures, and nothing leaves your device.
The words in this video
Prefer to read, or want to follow along while you watch? Here is everything the narrator says in this lesson, word for word.
Welcome back to AI Camp. Today, we get visual: how AI makes pictures and short videos — from nothing but your words. Words, you can do now. But sometimes a page needs a picture — and not everyone can draw, or find the right photo. Here is the idea. You describe a picture in plain words, and AI builds it — new, from the patterns it learned from millions of pictures. It does not fetch a photo. It makes one.
So we ask for the newsletter. Try this: a watercolor picture of a sunny community garden, tomato plants in neat rows, morning light. And back it comes, in seconds. A garden that never existed, painted from a sentence. That is the magic trick — and it really is a trick worth learning. But look closely. First tries are rarely right. Our fence leans, and one tomato plant thinks it is a tree. AI pictures often need a second look — and a second ask.
So we ask again — changing one thing at a time, like good editors do. Same garden, but in the late afternoon, with golden light, and a straight fence. Better. Same garden, warmer light, fence standing straight. Small changes to the words make big changes to the picture. Pictures like these are wonderful for cards, newsletters, and invitations — things that need charm more than they need a camera. This one is for Carol's birthday.
But hold on to one fact: an AI picture is not a photograph. Nobody stood in that garden. No camera clicked. It looks real — and it is made up. Learn the tells, and you will spot AI pictures anywhere: extra fingers, melted words, skin a little too shiny, shadows that fall the wrong way. Words are the weakest part. Ask for writing inside a picture and it often comes out melted. Keep the words out — add your own real words afterward.
And pictures can move. The same kind of ask can bring back a few seconds of video — the leaves stir, the light shifts, the garden breathes. Keep your hopes the right size: a few seconds of mood, not a whole film. A greeting, a header for a post, a moment of calm — that is what these clips do best. Now the honesty rule, and it matters. If AI made the picture, say so. Never pass a made-up picture off as a real photo — not of a place, not of an event, and never of a real person.
Today's camp rule: describe it clearly, check it closely — and say it was made with AI. Try it: describe a picture for something real in your life — a card, a newsletter, an invitation. Make two versions, changing one thing. Pick the better one, and if you share it, say it was made with AI. Next time: AI at work — meetings, planning, and follow-ups that practically handle themselves. Until then, keep imagining. See you at camp.