AI Image Prompt Builder
Choose the building blocks of your image and copy a prompt for your generator.
Runs in your browser. Nothing is uploaded.
How to build an image prompt
- Describe the subject: who or what is in the picture and what is happening.
- Pick a style, then add mood, lighting, camera, composition and colors. Every block is optional; skip the ones you do not care about.
- Choose an aspect ratio and anything to leave out.
- Pick the format your generator expects and copy the prompt.
The builder writes the prompt for you; it does not generate images. Paste the result into the image tool you already use.
Three output formats
- Sentences read like a short description: "A watercolor painting of a red fox in a snowy forest. Golden hour light." Tools that run on a language model, such as ChatGPT, Gemini, Copilot and Adobe Firefly, follow full sentences better than lists of keywords.
- --parameters gives a comma-separated prompt followed by
--arfor the aspect ratio and--nofor what to leave out, the syntax Midjourney uses. You can add--stylize,--chaosand--seed. The builder never adds a model version flag, so the prompt works with whichever version you select in your tool. - Tags is a comma-separated list for Stable Diffusion interfaces, with the negative prompt in its own box and a suggested image size of about one megapixel for the ratio you chose.
Choosing the blocks
- Style sets the medium: photo, film still, watercolor, pixel art, 3D render and more. The swatches next to each style are rough color hints drawn with CSS, not sample images.
- Lighting often changes a picture more than the style does. Golden hour and soft daylight flatter almost anything; low-key lighting and rim light add drama.
- Shot and lens borrow photography terms. A close-up with an 85mm lens and a shallow depth of field gives a classic portrait; a wide shot with a 24mm lens shows the setting.
- Composition and palette steer the layout and colors. Name your own colors if a brand palette matters.
- Negative prompt lists what you do not want: text, watermarks, extra fingers, a cluttered background.
Stuck? Surprise me picks a random style, mood, light and camera and keeps your subject.
Limits
Generators read the same words differently, and a longer prompt is not always better. Change one block at a time. Parameter ranges and support differ between tools and model versions, so check your tool's documentation if a parameter seems to be ignored.
Does this work for Midjourney, DALL·E, Stable Diffusion and Flux?
Yes. Use the parameters format for Midjourney, the sentences format for tools that run on a chat model, and the tags format with a negative prompt for Stable Diffusion. Flux reads plain sentences well.
Why is there no model version in the prompt?
Versions change often, and a flag for an old version can make your tool fall back to it. Choose the version in your tool's settings instead.
Can I use artist names?
You can type anything into the subject or details, but many tools limit prompts that copy a living artist's style. Describing the medium, era and technique usually works better.
Is anything I type stored?
No. The prompt is built in your browser. Only your last aspect ratio and output format are remembered in this browser.
What does --no do?
It tells Midjourney to avoid the listed things. Other tools use a separate negative prompt field, which is what the tags format gives you.
Images tools from our directory
Create stunning AI-generated images from text descriptions.
AI-driven image and video editing with natural language commands.
