This free AI image prompt generator turns a simple idea into a detailed AI image prompt, written the way each tool’s own docs recommend. Describe your subject, pick a style, lighting, camera, mood and colors, and you get ready-to-paste prompts for ChatGPT, Gemini (Nano Banana), Midjourney and Stable Diffusion, plus three variations to try.
Most disappointing AI images come from vague prompts. A good image prompt names the subject, the setting, the medium, the light and the framing, and says what to leave out. The builder fills all of that in for you, and nothing you type leaves your browser.
Free image prompt builder
Build an AI image prompt for any tool
Describe your picture once and get it written four ways: for ChatGPT, Gemini (Nano Banana), Midjourney and Stable Diffusion, each in the format its maker's docs recommend. It all runs in your browser, so nothing you type is sent anywhere.
Midjourney option: stylize
Style presets
One click sets the medium, style, light, angle, mood and colors. Your subject and details stay as they are.
Unlock all 50 style presets
Restyle any prompt in one click: film noir, claymation, risograph print, vintage travel poster, coloring page and 37 more, grouped by look.
Unlocked: all 50 presets are in the Style presets panel above. You are also on the weekly HOW TO // AI newsletter.
Free. You also get the HOW TO // AI newsletter once a week. Unsubscribe any time.
Have a picture you want to recreate? Try the image to prompt tool. Writing a text prompt instead? The AI prompt improver tightens it up.
Sources: OpenAI image prompting guide, Images in ChatGPT, Google Gemini image generation, Midjourney's Prompt Basics, Parameter List, No and Stylize pages, and the Stability AI API reference, checked October 18, 2026. Not affiliated with any of these companies.
How to use the image prompt builder
- Type your subject: who or what is in the picture. Add a setting if it matters.
- Pick a medium and style, or tap a style preset to set everything at once.
- Choose the lighting, camera, framing, mood and colors. Skip anything you don’t care about.
- Pick an aspect ratio and add anything to leave out, like blurry or extra fingers.
- Open the tab for your tool and copy the prompt. If the first image misses, try one of the variations.
What makes a good AI image prompt
Midjourney’s prompt guide lists the details worth deciding before you write: subject, medium, environment, lighting, color, mood and composition. The builder covers each one, plus the camera and lens for photographic looks.
- Be specific. Midjourney’s docs point out that “three cats” beats “cats”. In the same way, “warm golden hour sunlight” beats “nice light”.
- Say what you want, not what you don’t. Midjourney warns that if you mention a party with “no cake”, a cake might still appear. Use the leave-out box instead: the builder handles it the right way for each tool.
- Ask for photorealism by name. OpenAI’s guide says to request “photorealistic” or “real photograph” explicitly when that is the goal.
- Quote any words. Put text you want in the picture in quotes, keep it short, and check the spelling in the result.
- Describe the look in your own words. “Loose watercolor washes” or “film noir shadows” gets you a style without naming a living artist or a brand.
How each tool wants its prompt
| Tool | Prompt format | Leaving things out | Picture shape |
|---|---|---|---|
| ChatGPT Images | Plain sentences. For detailed requests, OpenAI suggests labeled sections for scene, subject, details and constraints | Say it in the prompt, for example “No text, logos or watermarks” | The aspect ratio picker, or say it in the prompt |
| Gemini (Nano Banana) | A descriptive paragraph. Google’s guide says to describe the scene in rich detail | Say it in the prompt | Say it in the prompt, for example “Aspect ratio 16:9.” |
| Midjourney | Short, comma-separated phrases, then parameters at the very end | –no followed by a comma-separated list | –ar, for example –ar 3:2 |
| Stable Diffusion (Stability AI API) | A descriptive prompt that clearly defines elements, colors and subjects | A separate negative prompt | aspect_ratio: 21:9, 16:9, 3:2, 5:4, 1:1, 4:5, 2:3, 9:16 or 9:21 |
One Midjourney trap to know: its moderation reads every word after –no on its own, so “–no modern clothing” is read as “no modern” and “no clothing”. Describe the clothing you do want instead. The builder warns you when this could happen.
Tips for better results
- Change one thing at a time. When an image is close, ask for one change and say what to keep. OpenAI’s guide suggests saying “change only” the thing you want changed, and Google recommends refining in the same conversation.
- Keep Midjourney prompts short. Its docs say short, simple prompts typically work best, so put the shape and exclusions into parameters.
- Use stylize on purpose. Midjourney’s –s runs from 0 to 1000 with a default of 100. Lower follows your words more closely, higher gets more artistic.
- Start from a preset. A preset sets the medium, light, angle, mood and colors in one click, so you can focus on the subject.
Writing a text prompt instead? Our free AI prompt improver rewrites a rough prompt into a clearer one.
Related on HOW TO // AI
- The complete guide to AI prompting: the habits behind every good prompt, written for people who aren’t engineers.
- ChatGPT vs Claude vs Gemini: which assistant to use for what.
- Gemini free vs Google AI Pro: what you actually get on each plan.
- Free AI tools: more builders, calculators and checkers like this one.
Checked against OpenAI’s image prompting guide and help center, Google’s Gemini image generation guide, Midjourney’s documentation and Stability AI’s API reference on October 18, 2026.
Frequently asked questions
What is a good prompt for AI image generation?
Name the subject and the setting, then the medium or style, the lighting, the camera angle or framing, the mood and the colors, and anything to leave out. Be specific, and describe what you want rather than what you don't want.
How do I write a Midjourney prompt?
Keep the description short and comma-separated, then add parameters at the very end, such as --ar 3:2 for the shape and --no for things to leave out. Midjourney's docs say to put a space before the dashes and no punctuation after the parameters.
How do I tell ChatGPT what to leave out of an image?
Write it into the prompt. OpenAI's image prompting guide says to state exclusions such as unwanted text, logos or watermarks. This builder writes them as a plain sentence for ChatGPT and Gemini, as --no for Midjourney and as a negative prompt for Stable Diffusion.
What is Nano Banana?
Nano Banana is Google's name for the Gemini image generation models. Google's guide recommends describing the scene in rich detail, so use the Gemini tab, which writes your prompt as one descriptive paragraph.
Can I still use DALL-E in ChatGPT?
OpenAI has retired the official DALL-E GPT from ChatGPT. Images are now made with ChatGPT Images, which OpenAI says is available on all tiers, and the ChatGPT tab in this builder is written for it.