How to Write AI Image Prompts (Examples That Actually Work)

The difference between a flat, generic AI image and a stunning one is almost always the prompt. The model can only draw what you describe — so the more clearly you describe it, the better the result. The good news is that good prompts follow a simple, repeatable structure. Here's how to write AI image prompts that actually work, with examples you can copy and adapt.

The anatomy of a good prompt

A reliable prompt usually answers five things, in this order:

1. Subject — what is in the image (a person, product, animal, scene).
2. Description & action — details and what the subject is doing.
3. Style — photo, oil painting, 3D render, anime, watercolor, etc.
4. Lighting & mood — golden hour, soft studio light, neon, moody.
5. Composition — close-up, wide shot, top-down, shallow depth of field.

Put together, a vague prompt like "a dog" becomes: "a golden retriever puppy sitting in tall grass, photorealistic, warm golden-hour light, shallow depth of field, close-up portrait." Same subject, completely different result.

The 5 building blocks, with examples

1. Subject

Be specific. "A woman" is weaker than "a young woman with curly red hair wearing a denim jacket."

2. Style

Style is the single biggest lever. The same subject as a photograph, an oil painting, or a 3D render looks entirely different. If you're not sure which words to use, see our list of 50+ AI art styles and the keywords that trigger them.

3. Lighting

Lighting sets the mood: "soft natural light," "dramatic rim lighting," "neon glow," "overcast." It's what makes an image feel cinematic instead of flat.

4. Composition

Tell the model how to frame it: "close-up," "wide establishing shot," "top-down flat lay," "rule of thirds," "shallow depth of field."

5. Detail & quality cues

Words like "highly detailed," "sharp focus," and "8k" nudge toward crispness — useful in moderation, but don't pile on twenty of them.

Copy-paste prompt examples

• Portrait: "studio portrait of a confident businesswoman, soft key light, neutral grey background, 85mm lens, shallow depth of field, photorealistic."
• Product: "a glass perfume bottle on a marble surface, soft diffused light, minimal, product photography, high detail, white background."
• Landscape: "misty pine forest at sunrise, volumetric light through the trees, wide cinematic shot, moody, ultra detailed."
• Art: "a fox in a snowy field, watercolor illustration, soft pastel palette, loose brush strokes."
• Logo / icon: "minimalist line-art mountain logo, single color, clean vector style, centered, plain background."

Common mistakes that ruin prompts

• Too vague. "Make it cool" tells the model nothing. Describe what "cool" looks like.
• Too many competing ideas. One clear scene beats five stacked concepts.
• Conflicting styles. "Photorealistic anime watercolor 3D render" cancels itself out — pick one.
• Endless quality words. A wall of "8k ultra hyper detailed masterpiece" adds little; clarity matters more.

How to iterate

Treat prompting like tuning, not gambling. Change one variable at a time — swap the lighting, then the style, then the composition — so you learn what each word actually does. Found a great image but want to tweak it? Send it to the AI image editor and adjust it with a follow-up instruction instead of starting over.

Master the five building blocks and you'll get usable images far more often. Start simple, add one detail at a time, and keep the prompts that work as templates for next time.