Introduction

AI image generators create pictures from text prompts (and sometimes from reference images). Tools change often—DALL·E, Midjourney, Stable Diffusion–based apps, Adobe Firefly, and school-approved alternatives all share a pattern: describe what you want; sample visual patterns the model learned; download or edit the result.

Building on Prompt Engineering, you will learn visual RTCCF: subject, style, camera/composition, and constraints. Fast, accurate typing on practice helps you adjust long visual prompts without losing details.

Learning Objectives

By the end of this lesson, you will be able to:

Describe text-to-image generation at a conceptual level
Write prompts with subject, setting, style, and technical cues
Iterate (vary seed/style/details) toward a usable image
Identify ethical red lines (likeness, deepfakes, trademark misuse)
Decide when AI art fits a school project—and when it does not

Main Lesson

How text-to-image roughly works

Models train on large sets of images paired with descriptions. At generation time they synthesize pixels (or latent representations) guided by your prompt. Outputs can look stunning yet:

Distort hands, text, logos, or maps
Invent impossible architecture
Reflect biases in training data (who looks “professional,” whose culture is centered)

Always inspect closely before publishing.

The visual prompt stack

Include:

Subject — Who/what is the focus?
Action / pose — What are they doing?
Setting — Where and when?
Style — Photo, watercolor, vector poster, claymation, blueprint…
Composition — Wide shot, close-up, rule of thirds, centered icon
Lighting / mood — Soft daylight, neon night, high-key classroom
Constraints — “No readable logos,” “school-safe,” “simple background for a poster”

Example:

Flat vector illustration of diverse students collaborating around a laptop in a bright classroom, clean shapes, soft teal and coral palette, wide shot, generous negative space on the left for title text, no logos, no watermarks.

Iterate like a designer

Change one variable at a time:

Switch style only (photo → risograph poster)
Adjust composition (close-up faces → wide room)
Simplify background for text overlays
Ask for “same scene, fewer objects”

Save versions with filenames that note the change (poster_v3_negspace).

Text inside images

AI often misspells words in pictures. For posters and slides:

Generate a background or illustration without words
Add titles in Canva, PowerPoint, Google Slides, or a design tool

That hybrid workflow looks more professional.

Ethics and policy

Do not use AI to:

Create non-consensual intimate or humiliating images of real people
Fake evidence (fake IDs, forged documents, false news photos)
Impersonate classmates or staff for bullying
Violate school rules or platform terms

Be careful with:

Celebrity / private individual likenesses
Trademarked mascots and brand logos
Claiming AI art as traditional photography when a contest forbids AI
Cultural or sacred symbols used out of context

Credit requirements vary. When allowed, label “AI-assisted image” honestly. Follow AI Ethics principles throughout.

Accessibility and inclusion

Avoid stereotypes in “default” characters—specify diversity intentionally when depicting people.
Ensure contrast if images sit behind text.
Provide alt text when you publish digital images.

When not to use AI images

Science diagrams that must be accurate (use teacher-approved figures)
Historical evidence (use archival sources)
Assignments that require your own drawing skill to be assessed
Anything needing legally cleared commercial brands without licenses

Key Definitions

Text-to-image — Generating pictures primarily from written prompts.
Style prompt — Language that steers medium, art movement, or rendering look.
Composition — Arrangement of subjects in the frame.
Upscale / refine — Tools that increase resolution or clean details (product-dependent).
Reference image — An uploaded picture used to guide composition or likeness (when allowed).
Deepfake risk — Harmful misuse of AI media to impersonate or deceive.
Hybrid workflow — Combining AI imagery with human layout, typography, and editing.
Alt text — Written description of an image for accessibility.

Examples

Example 1: Club flyer background

Generate abstract geometric shapes in school colors; add event text manually.

Example 2: Story illustration

Prompt a storybook watercolor scene matching a chapter mood; avoid depicting real students without consent.

Example 3: Mood board

Create four style variants (noir photo, comic panel, origami paper, isometric 3D) to choose a brand direction for a project.

Example 4: Icon set attempt

Ask for simple line icons on transparent backgrounds; manually fix inconsistent strokes in a vector tool if needed.

Real-World Scenarios

Scenario A — History fair

Riley wants a “photo” of a 1700s event. The teacher requires primary-source images. Riley uses archival prints instead and creates an AI timeline decoration only for the board’s border—clearly labeled.

Scenario B — Meme gone wrong

A group chat asks to AI-face-swap a classmate into a humiliating template. Jordan refuses, cites bullying policy, and suggests funny AI animals instead.

Scenario C — Pitch deck

A startup club generates product-mockup imagery, then replaces fake UI text with real screenshots before presenting to mentors.

Tips

Tip: Put the subject early in the prompt; overload of style words alone yields pretty but empty scenes.

Tip: For slides, leave negative space on purpose so titles remain readable.

Tip: Keep a swipe file of prompts that matched your school’s visual style guide.

Warnings

Warning: Generating harmful or explicit images of minors—or any non-consensual intimate imagery—is unethical and may be illegal. Do not attempt it.

Warning: Do not treat AI historical images as documentary evidence.

Did You Know

Did you know: Many platforms filter disallowed content, but filters are imperfect—judgment still matters.

Did you know: Same prompt ≠ same image every run; variation is normal in generative systems.

Common Mistakes

Expecting perfect spelling of poster headlines inside the image.
Crowding prompts with contradictory styles (“photo + oil painting + pixel art”).
Using real peers’ faces without consent.
Publishing biased stereotypes uncritically.
Skipping labels when contests or teachers require AI disclosure.

Interactive Exercise

Four Variants Challenge (15 minutes)

Create (or storyboard if tools blocked) four versions of one subject:

Photoreal classroom
Flat vector poster
Cut-paper craft look
Blueprint / technical line art

Write which fits a science fair vs. a poetry zine—and why.

Practice Questions

List five elements of a strong visual prompt.
Why add titles in a design tool instead of inside the AI image?
Name three unethical image uses.
When should you avoid AI images for schoolwork?
What is a hybrid workflow?

Mini Challenge

Produce one school-safe visual asset:

AI (or sketched plan) for background/illustration
Human-added title text
Alt text (1–2 sentences)
One-line AI disclosure if required

Summary

AI images are promptable draft visuals—not automatic truth or always-legal art. Specify subject, style, composition, and constraints; iterate carefully; finish typography yourself; and respect consent, culture, and policy. Ethical judgment travels with every download button.

Student Checklist

[ ] I can explain text-to-image basics
[ ] I wrote a multi-part visual prompt
[ ] I practiced iteration / variant thinking
[ ] I can list ethical red lines
[ ] I completed Four Variants
[ ] I finished practice questions and mini challenge

Teacher Notes

Prefer school-licensed tools; document allowed vs banned uses.
Critique bias: compare default “CEO” or “scientist” depictions.
Require disclosure labels on AI-assisted project visuals.
Offer offline alternative: collage or original sketches for equity when tools are restricted.
Connect to media literacy units.

FAQ

Q: Who owns AI images?

Rules depend on tool terms, local law, and school policy. Read licenses; when unsure, ask before commercial use.

Q: Can I use AI art in contests?

Only if rules allow. Many require disclosure or forbid AI entirely—read fine print.

Q: Why do hands look weird?

Models historically struggled with complex anatomy and counting fingers; improving but imperfect—edit or crop.

Q: What’s next?

Use AI as a coding assistant mindfully in AI for Coding Help.

Q: Typing?

Detailed visual prompts reward accurate keyboard skills—practice.

Related Lessons

Next Lesson CTA

You can brief an image model like a junior art director. Next, bring the same care to code: continue to AI for Coding Help and learn to ask for explanations, debug wisely, and keep learning yours.

AI Images