How-to · Design

How to brief an AI image tool

A brief you can reuse for Midjourney, Firefly, Ideogram, and other image generators so you spend credits on a job instead of on adjectives.

Illustration of a blank frame, a pencil, and simple color swatches in periwinkle, navy, and tangerine.
On this page

    Image generators reward a clear brief and punish a vibe. This is a brief you can paste above any prompt, whether you are in Midjourney, Adobe Firefly, Ideogram, or another generator from the design shortlist. Last reviewed October 2026.

    Write the brief before the prompt

    Fill these lines on paper or in a doc. Do not start in the tool.

    1. Job. What will the image be used for? A blog header, a concept for a client, a background behind type you will set yourself.
    2. Subject. Who or what is in the frame, in concrete nouns.
    3. Crop. Square, vertical story, wide banner. Name it now.
    4. Must not appear. Extra fingers are a cliché. Your list should include real constraints: no logos, no readable nonsense text, no real person’s face, no trademark.
    5. Reference rights. If you upload a photo, you need the right to upload it. If you do not, do not upload it.

    A mood sentence is allowed after those lines, not instead of them.

    Turn the brief into a prompt

    Use plain sentences. One possible shape:

    “A wide banner of a small bakery counter at morning, seen from the doorway, warm light, empty of people, no text, no logos, no watermark.”

    That prompt can fail, and you will know how it failed. “Cinematic award-winning hyper-real stunning” can also fail, and you will not know what to change.

    Avoid asking for a named living artist’s style. Describe lighting, materials, and era in words you would say to a photographer. Avoid asking for a celebrity or a private individual.

    Generate in a small batch

    Make four to eight images, not forty. Look at them against the brief, not against each other as a beauty contest. Keep one. Note which instruction was ignored. Change that instruction, not the whole mood. Stop when you have a usable crop or when you have spent the batch you allowed yourself.

    If words must appear inside the picture, Ideogram is the tool to try, and you still proofread every character. Often the better path is a picture with no text, and type set in Canva or an Adobe app where you control the spelling.

    Check rights before the file leaves your machine

    Read the tool’s current license for the plan you are on. Answer, in writing if this is client work:

    • Is commercial use allowed?
    • Did you upload anything you do not own?
    • Could a real person recognize themselves?

    If you hesitate, do not deliver. Tell the client you will use a photograph you can license, or a layout made of type and simple shapes. That conversation is cheaper than a takedown.

    File the winner with the brief

    Save the image next to the brief and the prompt. Future you will need to know what was asked, not only what was generated. When the tool’s terms change, the brief is also how you remember whether this file was ever cleared for ads.

    Questions this guide answers

    How long should a prompt be?

    Long enough to name the subject, the crop, the mood, and what must not appear. A paragraph of mood words without those four items wastes generations.

    Should I include brand names in the prompt?

    Do not ask a model to imitate a living artist's name or a trademark you do not own. Describe the look in plain visual terms.

    What if the words in the image are wrong?

    Treat rendered text as untrusted. Use Ideogram when lettering matters, then proofread, or set the type yourself in Canva or another editor.