Creating Pictures with AI

Type a description, get a picture. But how does that actually work — and why does it sometimes turn out weird?

Try it first

Prompt or Puzzle?

Below are 5 example prompts someone typed into an image-generation tool (these are illustrative examples, not a live image tool). For each one, guess: will it probably turn out clear, or get a little weird?

The prompt

Notice the pattern? Specific, familiar prompts tend to match. Vague prompts, text inside the image, unusual combinations, and tricky details like hands are where things get weird.

The idea

So how do text-to-image tools actually work?

A text-to-image tool has studied huge numbers of pictures, each paired with a caption describing it. Just like the chatbots from earlier lessons learned patterns in text, an image tool learns patterns connecting words to visual patterns — shapes, colors, textures, styles.

When you type a prompt, the tool isn't searching for an existing photo or pasting pieces of real pictures together. It's generating a brand-new picture — starting with a screen of fuzzy static and slowly bringing the visual patterns it learned into focus. That means even a very realistic-looking AI image isn't a real photo of anything that actually exists.

Byte says

The best prompts usually name three things: a subject (what), a setting (where), and a style or mood (how it should look). The more specific and familiar those are, the closer your result usually matches what you pictured.

Now try it for real

Create Your First AI Image

This activity happens in your own image generation tool — not on this page.

Before you start

Ask a grown-up before you open your image tool. Talk together about what's okay to ask it to draw — steer clear of real people you know, famous people, anything scary, or anything mean — and plan to review your pictures together before you save or share any of them.

1Open your image generation tool with a grown-up nearby.
2Write a first prompt. This is the same being-specific habit from Lesson 4 — try naming a subject, a setting, and a style or mood — for example: "a friendly orange fox, sitting in a snowy forest, in a soft watercolor style."
3Look at the result. Does it match what you pictured? What's different?
4Change one thing in your prompt — add a detail, fix something confusing, or try a different style — and generate again.
5Repeat step 4 until you've made 2–3 images total.
6Pick your favorite, and think about why you like it best.
7Review all your pictures with your grown-up before saving or sharing any of them.

Want to plan your first prompt before you open the tool? Draft it here.

Saved right in this browser, on this device — it's just for planning, and never sent anywhere.

Your turn

Your favorite picture

Once you've made your images, come back and answer: which one was your favorite, and what did you change between your first prompt and your best one?

This is saved right in this browser, on this device — it's never sent anywhere.

Recap

What you just learned

✓ Text-to-image AI generates a brand-new picture from learned patterns — it's not a real photo, and not existing photos pasted together.
✓ Specific prompts naming a subject, setting, and style usually match what you pictured more closely than vague ones.
✓ Some details — like hands and written text — are still tricky for AI, so weird results there are normal.
✓ Always review AI-generated pictures with a grown-up before saving or sharing them.

Show what you know

Quick Quiz

10 questions, picked at random from a bigger question bank — so if you take it again, you'll likely see a different mix. Let's go!

Up next

Lesson 7 · Storytelling with AI

You'll combine AI text and AI images to tell a story.

Go to Lesson 7 →