A good AI image prompt describes the subject, the setting, the style, the lighting and the composition in concrete words: “a red fox sitting in fresh snow at sunrise, soft golden light, shallow depth of field, wildlife photography”. You do not need a separate image prompt generator: the formula and the template below turn ChatGPT, Claude or Gemini into one.
The image prompt formula
Write the elements in roughly this order:
- Subject: who or what, with key details (age, color, material, action).
- Setting: where and when.
- Style or medium: photo, watercolor, 3D render, flat vector, oil painting.
- Lighting and mood: golden hour, studio light, neon, overcast, cozy.
- Composition: close-up, wide shot, top-down, centered, rule of thirds.
- Technical details if useful: lens, aspect ratio, color palette.
- Text in the image, in quotes, if any.
Example: “Close-up portrait of an elderly fisherman with a grey beard, wearing a yellow raincoat, on a harbor at dusk, cinematic photography, soft rim light, 85mm lens, muted blue palette.”
15 example prompts by style
| Style | Prompt |
|---|---|
| Product photo | “A white ceramic coffee mug on a light oak table, morning window light, minimalist studio photo, soft shadows” |
| Food | “Overhead shot of a bowl of ramen with soft-boiled egg and scallions, dark slate background, steam rising” |
| Portrait | “Young woman laughing in a sunflower field, golden hour, film photography, warm tones” |
| Architecture | “Modern glass house on a cliff over the sea, dusk, interior lights on, wide angle” |
| Watercolor | “Watercolor illustration of a Paris street café in spring, loose brush strokes, pastel colors” |
| Flat vector | “Flat vector illustration of a person working on a laptop with plants around, limited palette of teal and coral” |
| Isometric | “Isometric 3D illustration of a tiny coworking office, soft lighting, clean pastel colors” |
| Logo concept | “Simple geometric logo of a fox head, single color, on white background” |
| Poster with text | “Retro travel poster of the Alps with the words "Visit Switzerland" in bold Art Deco lettering” |
| Fantasy | “Ancient library inside a giant tree, glowing lanterns, mist, detailed digital painting” |
| Sci-fi | “Cyberpunk city street at night in the rain, neon signs in Japanese, reflections on the pavement” |
| Children’s book | “Cute hedgehog reading a book under a mushroom, children’s book illustration, soft colored pencil” |
| Infographic style | “Clean infographic-style illustration of the water cycle with simple icons, white background” |
| Interior design | “Scandinavian living room with a grey sofa, large plants, natural light, interior design photo” |
| Social media banner | “Wide banner of abstract flowing gradients in purple and blue, space on the left for text, 16:9” |
A template that turns a chatbot into an image prompt generator
Paste this into ChatGPT, Claude or Gemini:
You are an expert at writing prompts for AI image generators. I will describe an image idea in plain words. Write three prompt versions: one photographic, one illustrated, one in a style of your choice. Each prompt should name the subject, setting, style, lighting, composition and color palette in under 60 words. If my idea is vague, ask me two questions first. My idea: [describe your image]
Add “format the prompts for Midjourney” or “for Stable Diffusion with a negative prompt” if you use those tools.
Tips for specific generators
ChatGPT and Gemini understand long natural sentences and can edit an image through follow-up messages (“make the background darker”). Midjourney rewards strong style words and supports parameters such as aspect ratio. Stable Diffusion and Flux interfaces often support negative prompts (what to exclude) and weighting. Ideogram and the latest GPT image models are the most reliable at rendering text inside images. See the best AI image generators for a full comparison.
Common mistakes
Vague adjectives (“beautiful”, “amazing”) add little. Too many unrelated details confuse the composition. Asking for famous people or trademarked characters often triggers refusals. And contradicting instructions (“minimalist, highly detailed”) produce muddy results.
To reverse the process and get a prompt from an existing picture, read image to prompt AI. For text prompts in general, see how to write prompts.
Frequently asked questions
What makes a good AI image prompt?
A clear subject, the setting, the style or medium, lighting, composition and any text that must appear. Concrete visual details work better than abstract adjectives like 'amazing'.
Do image prompts work the same in every generator?
Mostly. ChatGPT and Gemini understand full sentences well, Midjourney responds strongly to style words and its own parameters, and Stable Diffusion or Flux interfaces may support negative prompts and weights.
Can ChatGPT write image prompts for Midjourney?
Yes. Describe the image you want and ask ChatGPT, Claude or Gemini to write a Midjourney-style prompt; the template in this article does exactly that.