cd /news/generative-ai/6-design-concepts-for-better-ai-imag… · home › topics › generative-ai › article
[ARTICLE · art-161695] src=gist.github.com ↗ pub= topic=generative-ai verified=true sentiment=↑ positive

6 design concepts for better AI images: before/after prompts, explanations, and reusable patterns

A developer published a set of six design concepts and twelve before/after prompts for getting more controllable AI image generation, arguing that vague descriptors like "beautiful" or "eye-catching" leave visual decisions to the model. The revised prompts instead specify dominant elements, type hierarchy, negative space, limited palettes and concrete visual metaphors, with guidance on checking results at thumbnail size. The examples come from the developer's Better AI Images video.

by read6 min views1 publishedOct 31, 2026

Six design concepts, twelve before/after prompts, and what to look for in the results.

“Beautiful,” “professional,” and “eye-catching” tell an image model how you want to feel about the result. They leave a lot of the visual decisions open.

More useful questions: What should someone notice first? What should the empty space do? How can the composition explain the idea? What shapes should communicate the personality?

The examples below are from my Better AI Images video. Use them as a small design vocabulary you can borrow for your own work.

The job: Make the most important information the easiest thing to notice.

Before prompt

A poster for AFTER DARK, a rooftop roller-skating party. Saturday, 8 PM. Make it fun and eye-catching.

After prompt

A typographic poster for AFTER DARK, a rooftop roller-skating party. Saturday, 8 PM. Oversized cropped lettering, an asymmetric grid, clear type hierarchy, generous negative space, cobalt and pink.

What changes in the example: The busy neon poster becomes a composition dominated by the AFTER DARK lettering, with the smaller event details arranged around it.

Why this gives you more control: “Fun and eye-catching” leaves the model to decide what gets attention. The revised prompt establishes a dominant element, a layout structure, and breathing room. It gives the title a clear role instead of asking every element to compete.

Borrow this: “Make [the main message] dominant, group secondary details, and leave generous negative space.”

Check the result: At thumbnail size, what do you read first? Is it the thing you intended? Cropped lettering should still communicate the title.

The job: Give the empty area a role in the image.

Before prompt

A book cover for OFFLINE featuring a phone and a forest. Make it striking and sophisticated. Use only the title.

After prompt

A negative-space book cover for OFFLINE: a phone silhouette whose screen opens into a forest path. Forest green and cream, bold flat shapes, restrained typography. Use only the title.

What changes in the example: A dimensional phone-in-a-landscape scene becomes a graphic cover where the phone outline and forest opening work together.

Why this gives you more control: The revised prompt describes a relationship between the objects. The opening inside the phone becomes part of the idea, rather than asking for a phone and a forest to simply appear in the same picture. The limited palette and simpler shapes help that relationship read.

Borrow this: “Use negative space so [one shape] reveals [another shape or idea].” Check the result: Can you recognize the relationship quickly, or does it need an explanation? Negative space can also simply provide breathing room; it does not always need to hide a second image.

The job: Turn an abstract feeling into a concrete visual situation.

Before prompt

An illustration about digital overload for a technology newsletter. Include a person and phone notifications. Make it engaging. No text.

After prompt

An editorial visual metaphor for digital overload: phone notifications become a rising tide around a person standing on a tiny island. Bold geometric shapes, exaggerated scale, lavender and cobalt. No text.

What changes in the example: A person surrounded by notifications becomes a person on an island surrounded by a sea of notifications.

Why this gives you more control: The rising tide expresses the pressure of being overwhelmed. The model receives a specific visual mechanism for the feeling, rather than another generic “stressed person with a phone” scene.

Borrow this: “Show [abstract idea] as [specific physical situation].” Check the result: Does the metaphor communicate the intended feeling? The actual output here is dimensional, even though the prompt asks for geometric shapes. Evaluate the picture you got, not just the words you supplied.

The job: Show how the parts of a place or system fit together.

Before prompt

An illustration for a website about running a small online business from home. Include a workspace, products, and deliveries. No text.

After prompt

An isometric cutaway illustration of a home business: workspace upstairs, product storage below, delivery station outside. Consistent parallel axes, simplified geometry, pastel color blocks, clean white background. No text.

What changes in the example: A home-working scene becomes a building cutaway showing the workspace, storage, and delivery area in one view.

Why this gives you more control: “Isometric” specifies the viewing approach, while “cutaway” makes the interior visible. Naming where each activity happens gives the composition an explanatory structure. This can be useful when a pretty room alone does not tell the whole story.

Borrow this: “An isometric cutaway showing [area A], [area B], and how they connect.” Check the result: Do the floors, stairs, walls, and scale make sense? Treat it as an explanatory illustration; verify details separately if they need to be technically accurate.

The job: Make the letters themselves carry the visual idea.

Before prompt

A poster for BLOOM, a weekend flower market. Include flowers. Make it fresh and eye-catching. Only text: BLOOM / FLOWER MARKET / SUNDAY 10 AM TO 4 PM.

After prompt

A typographic illustration poster for BLOOM: the title's readable letterforms themselves grow into stems, leaves, and blossoms. Bold flat vector shapes, raspberry and pale pink on forest green. Only text: BLOOM / FLOWER MARKET / SUNDAY 10 AM TO 4 PM.

What changes in the example: Flowers surrounding a title become floral forms built into the BLOOM lettering.

Why this gives you more control: The prompt specifies where the concept belongs: in the letterforms. It connects the subject and typography instead of using flowers only as decoration around otherwise conventional text.

Borrow this: “Readable letterforms that become [a visual property of the subject].”

Check the result: Can you still read the word? Check spelling, dates, and event details before using the image. The complete poster matters here; an extreme flower crop can hide the typographic idea.

The job: Give a character's personality a visible shape.

Before prompt

A dragon mascot for a beginner-friendly coding app. Make it welcoming and memorable. White background, no text.

After prompt

A dragon mascot for a beginner-friendly coding app, using rounded shape language: pear-shaped body, oversized head, short blunt horns, tiny wings, thick curled tail. Flat vector illustration, bold silhouette, cobalt and pink. White background, no text.

What changes in the example: A detailed dimensional dragon becomes a simplified cobalt-and-pink mascot with deliberately rounded proportions.

Why this gives you more control: “Welcoming” is an intention. Body shape, head size, horns, wings, and silhouette are concrete decisions the model can depict. Rounded proportions support the approachable character intended here, though expression, pose, and context also matter.

Borrow this: “Use [shape language], with [specific proportions], to support [intended personality].” Check the result: Does the silhouette work small? Does the character feel right for the audience? This pair also changes palette and rendering style; the difference is not caused by proportions alone.

Before adding more adjectives, decide:

  1. What is this image supposed to do?
  2. What should someone notice first?
  3. What visual relationship would make the idea clearer?
  4. Which layout, shape, or typography decisions can I name?
  5. What should I check before calling the result finished?

A useful prompt structure:

Create [format] for [audience and purpose]. Make [main message or subject] dominant. Use [specific layout, visual relationship, or shape decisions]. Keep [important constraints]. These are illustrative before/after examples, not a controlled experiment. The revised prompts often change several things at once, and wording does not guarantee the same output on a different run or image model. The useful habit is to make visual decisions deliberately, inspect the result, and revise what is still unclear.

Start with one concept that solves your actual problem. You do not need to cram all six into every prompt.

── more in #generative-ai 4 stories · sorted by recency
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/6-design-concepts-fo…] indexed:0 read:6min 2026-10-31 · —