15 of the Best AI Image Prompts (and How to Write Your Own)

The best AI image prompts are specific: they name a subject, what it is doing, the setting, the style or medium, the lighting, and the framing, in that rough order. "A dog" gives you a generic stock photo. "A scruffy terrier puppy asleep on a stack of library books, warm window light, shallow depth of field, 35mm film photo" gives you the picture you meant. Below are 15 ready-to-use prompts you can copy, plus the formula behind them so you can write your own.

The formula behind every good image prompt

Almost every strong prompt answers six questions. You do not need all six every time, but the more you leave out, the more the AI fills in with something average.

1. Subject. Who or what, with one or two defining details. "An elderly fisherman with a white beard," not "a man."
2. Action or pose. "Mending a net," "laughing mid-turn," "looking straight at the camera."
3. Setting. "On a wooden dock at dawn," "in a cluttered 1970s kitchen."
4. Style or medium. "Photorealistic," "watercolor," "3D render," "flat vector illustration," "anime," "oil painting."
5. Lighting and mood. "Soft overcast light," "golden hour," "neon glow," "moody and dark," "bright and cheerful."
6. Framing. "Close-up portrait," "wide shot," "overhead view," "shot on a 50mm lens."

Put the most important idea first. Many image models give more weight to early words, so if the color of a jacket matters most, do not bury it at the end of a long sentence.

15 of the best AI image prompts by category

Swap the details in brackets for your own. Each one is built on the formula above.

Portraits (for men, women, or profile pictures)

1. Close-up portrait of [a man in his 30s with a short beard], wearing a charcoal wool coat, standing on a rainy city street at night, neon reflections on wet pavement, cinematic lighting, shallow depth of field, photorealistic
2. Portrait of [a woman with curly auburn hair] laughing, sitting in a sunlit cafe, soft natural window light, warm tones, 85mm lens, candid lifestyle photography
3. Professional headshot of [a person in a navy blazer], neutral light gray background, soft even studio lighting, sharp focus on the eyes, friendly expression
4. Stylized 3D cartoon avatar of [a teenage boy with messy black hair and headphones], bright colors, soft shading, simple gradient background, centered head and shoulders

Fantasy and characters

5. A female elf warrior in detailed silver armor, long flowing white hair, holding a glowing staff, ancient forest background, dramatic rim lighting, digital painting, highly detailed
6. A steampunk inventor with brass goggles covered in tiny gears, Victorian clothing, a small mechanical owl made of copper on his shoulder, cluttered workshop, warm lamplight

Landscapes and scenes

7. Snow-capped mountains above a misty pine valley at night, vibrant aurora borealis overhead, starry sky, wide landscape shot, photorealistic
8. A cozy cabin in a snowy forest at dusk, smoke rising from the chimney, warm light glowing in the windows, a deer near a frozen stream, storybook illustration style
9. A futuristic city street at night, flying cars, holographic shop signs, light rain, cyberpunk style, low camera angle

Products and practical images

10. A ceramic coffee mug on a white marble counter, morning sunlight from the left, soft shadows, minimalist product photography, plenty of empty space on the right
11. Flat lay of a healthy breakfast bowl with berries and granola on a linen tablecloth, overhead view, bright natural light, food photography

Art styles

12. A bioluminescent jellyfish drifting through a deep sea cave, ethereal blue glow, dramatic shadows, digital art
13. A quiet Paris street corner in spring, loose watercolor painting, soft pastel colors, visible paper texture
14. A red fox curled up asleep in autumn leaves, children's book illustration, gentle colors, simple shapes
15. Retro travel poster of a lighthouse on a rocky coast, art deco style, bold flat colors, strong geometric shapes

How to write prompts for ChatGPT, Gemini and other image tools

The formula works everywhere, but each tool reads prompts a little differently.

• ChatGPT and Gemini understand normal sentences well. You can write "Make the background darker and move the cat to the left" as a follow-up, and they will usually edit rather than start over. Describe what you want in plain language, and say what you do not want in words ("no text in the image").
• Art generators that use parameters (several popular Discord and web tools work this way) respond well to short, comma-separated descriptive phrases, with settings added at the end. The most useful one is --ar for aspect ratio, such as --ar 16:9 for a wide image or --ar 2:3 for a portrait. --no excludes things, for example --no text, watermark.
• Stable Diffusion tools usually have a separate negative prompt box. Be specific there: "blurry, low resolution, extra fingers, deformed hands, watermark" works better than a single vague word.

Features and syntax change often, so if a parameter does nothing, check the tool's current help page.

A worked example: from flat to finished

Sam wants a picture of a coffee shop for a small business newsletter. Her first prompt is coffee shop. She gets a generic, slightly plastic interior that could be anywhere.

She runs through the formula:

• Subject: a small neighborhood coffee shop
• Action: a barista pouring latte art
• Setting: exposed brick walls, plants in the window
• Style: candid photography
• Lighting: early morning sun through the front window
• Framing: medium shot from behind the counter

New prompt: A barista pouring latte art in a small neighborhood coffee shop, exposed brick walls, plants in the front window, early morning sunlight streaming in, candid photography, medium shot from behind the counter, warm tones

This version is close. The barista's hands look odd, so instead of rewriting everything she changes one thing: she adds "hands in sharp focus, natural hand pose." The third result goes in the newsletter. Three tries, about five minutes, and each change was small and deliberate.

That is the habit worth building: change one or two details per attempt, so you can tell what actually made the difference.

Common prompt mistakes and how to fix them

• Being too vague. "A beautiful landscape" means nothing specific. Name the place, time of day, weather and style.
• Stuffing in too much. Ten subjects and five styles confuse the model. If a scene is complex, build it in stages or generate pieces separately.
• Contradicting yourself. "Minimalist, highly detailed, busy background" pulls in opposite directions. Pick one direction.
• Rewriting from scratch every time. When an image is almost right, keep the prompt and change a single phrase. Starting over throws away what worked.
• Ignoring aspect ratio. A portrait in a wide frame wastes space. Set the shape before you start.
• Copying artists or brands. Leave out real logos, copyrighted characters and living artists' names. Describe the qualities you like instead ("bold ink lines, muted earth tones").
• Expecting readable text. Many models still garble words. Add text afterward in an editor when it matters.

Common questions

Where can I find good AI image prompt examples?

Prompt galleries and community forums such as Reddit are full of examples, and many tools show the prompt behind featured images. Treat them as starting points: copy the structure, then swap in your own subject and details.

Are the best prompts for ChatGPT different from other image generators?

The core is the same: subject, action, setting, style, lighting, framing. ChatGPT handles full sentences and conversational edits well, while parameter-based art generators tend to reward tight descriptive phrases and parameters like --ar.

How long should an AI image prompt be?

Usually one to three sentences, or about 20 to 60 words. Shorter prompts leave too much to chance, and very long ones tend to have details ignored. Put the most important details first.

Why does the AI keep ignoring part of my prompt?

The detail is probably buried late in the prompt or competing with something else. Move it to the front, state it more plainly, and remove anything that contradicts it.