Skip to content

AI tool

AI Prompt Generator

Describe what you want, choose the model you will actually run it in, and get a prompt written to that model's conventions — with a matching negative prompt.

Describe the image

Write it however you would explain it to a person. The formatting for Midjourney is handled for you.

0 / 1,200

What is in the shot, and what is it for?

Prompt options

Which model the wording is written for, and how far the description gets pushed.

Comma-separated descriptors with parameter flags

Fill in lighting, composition and mood that the idea implies.

Added as an --ar flag.

Optional. Leave unset to let the description drive the style.

Recent prompts

Saved in this browser only. Never uploaded anywhere.

About the prompt generator

Describe what you want in plain language, choose the model you are actually going to run it in, and get a prompt written to that model's conventions — plus a matching negative prompt where the model supports one.

The reason to pick the model up front is that the same words behave differently in each. Midjourney wants dense comma-separated descriptors and trailing flags; Flux follows long descriptive prose literally; Ideogram wants the text you need rendered stated first. One generic prompt served to all of them is a prompt tuned for none of them.

When this is the right tool

  • You know what you want to see but your prompts keep producing something adjacent to it.
  • You are switching between models and the phrasing that worked in one is failing in another.
  • You want a negative prompt that addresses real recurring faults rather than a copied list.
  • You have a reference image instead of a description — use image to prompt.

How to use it

  1. Describe the image

    Plain language is fine and preferred. Subject first, then what it is doing, then where — the structure can be tidied for you, but the content has to come from you.

  2. Choose your model

    ChatGPT Images, Midjourney, Flux, Ideogram or Adobe Firefly. The syntax, the length and whether a negative prompt is produced all follow from this choice.

  3. Set aspect ratio and style

    The ratio is emitted in whatever form the chosen model expects — a flag for Midjourney, plain guidance elsewhere.

  4. Generate, then iterate

    Change one clause at a time when a result is close, rather than rewriting the whole prompt and losing the parts that worked.

Common questions

Which models can it write for?

Midjourney, Flux, Ideogram, Adobe Firefly and ChatGPT Images. Each has its own rules — Midjourney wants comma-separated descriptors and parameter flags, Flux wants long literal prose, Ideogram leads with the text to render, Firefly wants something short and commercially safe — and the prompt is written to whichever you pick.

What is a negative prompt, and does every model take one?

It lists what you do not want in the image. Midjourney, Flux and Ideogram support one; Midjourney receives it as --no. ChatGPT Images and Firefly have no negative field, so for those the exclusions are folded into the prompt itself.

Can I set the aspect ratio and style?

Yes. Aspect ratio, style preset and rendering approach are all controls, and they are expressed in whatever way the target model expects — as a flag where the model takes flags, as description where it does not.

Are my prompts saved anywhere?

The last 20 stay in your browser's localStorage so you can pull one back up, and nothing else. There is no account and no server-side history; clearing your browser data removes them.

Do I need an API key?

Yes — a free Gemini key from Google AI Studio, pasted once in Settings. It is stored on your device and used only to write your prompts. No billing account is required.

Can I use the prompts commercially?

The prompt text is yours to use. Whether the resulting image can be sold or submitted to a stock agency depends on the generator's own terms and the agency's AI policy, which is a separate question from the prompt.

Guides on this

Longer answers to the questions this tool raises.

Related tools