How to Write Prompts for AI Image Generation

Why prompts matter so much

With AI text tools, a vague question often still produces a useful answer. AI image tools are less forgiving. A vague prompt tends to produce a generic result, and generic AI images look generic. They have a certain sameness that makes them easy to spot and easy to ignore.

A strong prompt produces something specific, on-brand, and useful. The difference between a weak and strong prompt is almost entirely about how clearly you describe what you want.

This tutorial gives you a practical formula and a vocabulary you can start using immediately.

The prompt formula

Think of an image prompt in five parts. You do not need to use all five every time, but having all five in mind helps you build richer descriptions.

Subject: What is in the image? Be specific. Instead of "a person," try "a woman in her 30s sitting at a bright modern desk, focused on a laptop." Instead of "a city," try "a narrow street in a European old town at dusk with warm cafe light spilling onto the cobblestones."

Style: What visual style should it have? This is one of the most powerful levers in your prompt. Examples: photorealistic, cinematic photograph, flat design illustration, watercolor painting, minimalist, editorial magazine style, vintage poster, 3D render.

Composition: How is the image framed? Examples: close-up portrait, wide establishing shot, bird's eye view, centered symmetrical, rule of thirds.

Lighting: Lighting dramatically changes the feel of an image. Examples: golden hour sunlight, soft studio lighting, moody overcast, harsh midday sun, dramatic side lighting, bright and airy, neon glow.

Mood: What feeling should the image convey? Examples: calm and minimal, energetic and vibrant, serious and professional, warm and inviting, tense and dramatic.

Comparing weak and strong prompts

Here is the same idea expressed as a weak prompt and a strong prompt.

Weak: "A woman working."

Strong: "A woman in her early 30s working at a standing desk in a bright, modern home office. Natural light from a large window. Soft shadows. Minimal, Scandinavian interior. Photorealistic. Warm and focused mood."

The second prompt gives the AI everything it needs: who, where, what the environment looks like, the lighting, the style, the mood. The result will be far more specific and usable.

Weak: "A logo for a coffee shop."

Strong: "A minimalist logo concept for an artisan coffee shop. Simple line art of a coffee cup with steam forming a leaf shape. Black on white. Clean, modern, no text."

The second version specifies the style, the visual elements, the treatment, and the format.

Style vocabulary worth knowing

Building a mental library of style words pays dividends every time you generate an image. Here are categories to draw from:

Photography styles: photorealistic, editorial photography, portrait photography, street photography, product photography, macro photography, documentary style.

Illustration styles: flat design, line art, vector illustration, watercolor, ink wash, gouache painting, pencil sketch, vintage illustration, retro poster.

3D and render styles: 3D render, isometric illustration, clay render, octane render, blender 3D.

Mood and atmosphere: cinematic, moody, dramatic, ethereal, minimalist, maximalist, vibrant, desaturated, high contrast, soft and dreamy.

Lighting descriptors: golden hour, blue hour, studio lighting, soft box, hard light, backlit, overcast, neon lit, candlelit.

Aspect ratios

Most AI image tools let you specify the shape of your output. Always choose the aspect ratio that matches your intended use before you generate.

Common ratios and their uses:

• 1:1 (square) for Instagram posts and profile images
• 16:9 (landscape) for YouTube thumbnails, presentations, website banners
• 9:16 (portrait) for Instagram Stories, TikTok, and Reels
• 4:3 for general landscape use and older presentation formats
• 3:2 for standard photography proportions

Generating at the right ratio from the start saves you from awkward cropping.

What to avoid in prompts

A few common mistakes make prompts less effective.

Being abstract when you could be concrete. "Beautiful landscape" is less useful than "a mountain valley in autumn, orange and red foliage, a river in the foreground, misty morning atmosphere."

Stacking too many contradictory ideas. Asking for "modern and vintage and minimalist and detailed" pulls the AI in multiple directions. Choose a direction.

Describing emotions directly. Instead of "a happy scene," describe the things that create happiness: warm light, smiling faces, bright colors.

Forgetting the format. Not specifying a style often produces something generic. Even adding "photorealistic" or "flat illustration" makes a significant difference.

A prompt to start with

When you are not sure where to begin, use this structure:

[Subject], [setting or background], [style], [lighting], [mood], [aspect ratio]

For example:

A small business owner reviewing analytics on a laptop, modern minimalist office, editorial photography style, soft natural window light, calm and focused mood, 16:9

Paste this into any AI image tool and adjust from there. The next tutorial covers what to do when you want to refine the results you get.