How We Write Image Prompts That Build Better Slides

My slides are mostly pictures. A good image can carry an idea faster than a paragraph of text, but only if the image actually shows the idea. Early on, most of what I got from AI image tools was generic clip art that looked fine and taught nothing. The fix wasn't a better tool. It was writing better prompts.

Here's how we do it. I give Claude a single bullet point from my slide outline, and it writes a detailed prompt for Google's Nano Banana image tool. I run the prompt, drop the image into the slide, and move on. Below are the rules we follow and three real examples from a unit on primary and general elections.

Lock in one style and repeat it word for word. Every prompt in a deck starts with the same style description: "A comic-book style illustration, halftone shading, bold black outlines, bright comic color palette, matching the same visual style as the previous slides." If you describe the style differently each time, every slide looks like a different artist drew it. Copying the exact same opening keeps the whole deck consistent.

Describe a comic panel, not a photo. Tell the AI the layout first, whether that's a single scene, two side-by-side panels, or a room split down the middle. Then describe the characters with simple, specific poses and expressions. Words like "excited" or "dramatic" are too vague to draw.

Spell out every word of text. If you want a caption in the image, write the exact words in quotes, such as: a bold comic caption banner reads 'YOU HAVE TO QUALIFY BEFORE YOU COMPETE.' Vague instructions like "add a title" produce garbled letters.

Let contrast do the teaching. The strongest slides put two things side by side and make the difference obvious through size, color, lighting, and energy. When the picture shows the contrast, the caption barely has to explain it.

Example 1: Explaining an idea with an analogy. The concept was that a primary is a qualifying round before the general election. A sports comparison made it click.

"A comic-book style illustration, wide horizontal scene split into two side-by-side panels by a thick black divider, halftone shading, bold black outlines, bright comic color palette, simple and easy to read for a beginner classroom audience.

LEFT PANEL, labeled 'TRYOUTS' in bold comic lettering: A high school gym scene. A coach with a clipboard and whistle stands watching a group of five cartoon athletes in plain gray practice jerseys running drills, sweating, giving their best effort. The coach's hand points at one standout athlete who's clearly performing best — comic motion lines and a small spotlight beam highlighting them. A sign on the gym wall reads 'MAKE THE TEAM.'

RIGHT PANEL, labeled 'GAME DAY' in bold comic lettering: The same standout athlete, now wearing a full team jersey and number, running onto a packed stadium field under bright lights, crowd cheering in the stands, confetti in the air. A scoreboard in the background reads 'HOME vs. AWAY.'

Above both panels, a single bold comic caption banner reads: 'YOU HAVE TO QUALIFY BEFORE YOU COMPETE.'

Below the caption, a smaller subtitle line reads: 'Tryouts pick the player. The season decides the winner.'

Overall style: same bold-outline, halftone-dot comic aesthetic as the previous slides, clean and immediately understandable to a high school audience."

Example 2: Making the imbalance visible. The idea was that candidates often respond to a small, loud group instead of the larger, quieter majority. Instead of explaining it, the image shows it through size and color.

"A comic-book style illustration, single wide scene, halftone shading, bold black outlines, bright comic color palette, matching the same visual style as the previous slides.

The scene shows a large circular crowd viewed from a slightly elevated angle, like a stadium floor. In the very center of the crowd stands a small tight cluster of about six loud, exaggerated supporters — fists raised, mouths wide open shouting, oversized signs, drawn much bigger and more energetic than everyone else, almost bursting out of the panel with motion lines and small starburst shapes around them. Standing on a small box in the middle of that cluster is the candidate, arms raised triumphantly, clearly reacting to and celebrating with this small loud group.

Surrounding that small loud cluster, filling the rest of the huge crowd, are dozens and dozens of ordinary cartoon people drawn calm and neutral — arms crossed or hands in pockets, plain neutral expressions, no signs, just standing quietly. They're drawn smaller in detail and duller in color compared to the vivid, oversized loud cluster, even though there are far more of them.

Above the scene, a bold comic caption banner reads: 'A LOUD MINORITY. NOT THE MAJORITY.'

Below it, a smaller subtitle line reads: 'The candidate is reacting to the loudest voices in the room — not most of the room.'

Overall style: same bold-outline, halftone-dot comic aesthetic as the previous slides, the visual size and color contrast between the small loud cluster and the large quiet crowd should make the imbalance immediately obvious at a glance."

Example 3: Fixing the extra-arms problem. This one taught us the most. The first version of this prompt asked one character to hold a megaphone, gesture wildly, and turn away all at once, while the other character held out a folder. The result had people with three or four arms. AI image tools get confused when you describe hands doing different things at the same time. The fix was to give each arm one simple job and add a line about anatomy.

"A comic-book style illustration, single wide scene set at a shared podium or negotiating table, halftone shading, bold black outlines, bright comic color palette, matching the same visual style as the previous slides. Each character has exactly two arms and two legs, simple clean anatomy, no extra limbs.

On the left, a wildly exaggerated cartoon candidate stands facing away from the other character, toward an unseen cheering crowd off-panel. He holds a megaphone up to his mouth with one hand; his other arm hangs relaxed at his side. Wide-eyed manic grin, hair sticking up, comic action lines and sparkles around him. A speech bubble reads 'AND THAT'S WHY WE SHOULD—' trailing off mid-sentence, ignored.

On the right, a calm, reasonably-dressed cartoon politician in a plain suit stands with arms crossed over his chest (both arms folded, not reaching), visibly frustrated — one eyebrow raised, a single sweat drop, a small storm cloud with a lightning bolt above his head. A folder labeled 'REAL SOLUTIONS' sits on the table in front of him, untouched, rather than being held out — so no extra reaching arm is needed. A thought bubble above him shows a simple cartoon handshake, contrasting with what's actually happening.

Between them, a small microphone stand sits unused and slightly knocked sideways.

Above the scene, a bold comic caption banner reads: 'TALKING PAST EACH OTHER.'

Below it, a smaller subtitle line reads: 'Extreme candidates play to the crowd — not to solving the problem.'

Overall style: same bold-outline, halftone-dot comic aesthetic as the previous slides, simple two-armed poses for both characters, strong contrast in body language between chaotic motion on one side and rigid stillness on the other."

One more lesson: skip "transparent background." When we asked for a transparent background, Nano Banana drew the gray-and-white checkerboard pattern into the image instead of leaving the background empty. A solid white background works better and blends into most slides.

Build the deck in two passes. I build a plain, text-only version of the deck first, with short phrases on each slide. Once the structure is right, we go back and write an image prompt for each slide. Separating the two steps keeps me from redesigning slides around images before I know what the lesson needs.

You can copy any of these prompts and swap in your own subject. Keep the style opening the same, describe the layout, spell out the text, and let the contrast in the picture do the explaining.