How to write a useful
MOSS-Image prompt
Start with the picture you need. Describe the subject and setting, separate the words that belong in the image, and decide how much control you need over the final layout.
1. Make the brief specific
“Make a food poster” leaves most of the decisions open. A better brief names the food, the view and the purpose: one double smash burger with melted cheddar and fries, photographed for a neighborhood restaurant lunch promotion.
Add details that distinguish your idea. If there should be one item, say one. If the background should stay simple, say so. Include anything that must not appear, such as prices or invented brand names. A useful prompt should reflect those decisions.
2. Keep visible words separate
Instructions and image text serve different purposes. “Put a headline at the top” describes placement. “MAKE IT A BURGER DAY” is the headline itself. Enter the exact characters in the words field, including punctuation and line breaks.
IMAGE BRIEF An Instagram lunch promotion showing a double smash burger and crispy fries. Warm diner food photography. No prices. WORDS IN THE IMAGE MAKE IT A BURGER DAY Smash burgers. Fresh fries. COMPOSITION Headline at the top. Food in the lower half. A clear gap between the two.
This separation makes the intended result easier to review. It does not guarantee that an image model will render every character correctly. Check the finished image, especially names, numbers and smaller supporting text.
3. Describe relationships and space
Composition is easier to follow when it explains where subjects sit in relation to one another. Try “the teapot is centered, with its handle on the right” or “the shop sits on the left, with open pavement on the right.”
For a poster, think about the available space before asking for more detail. A large product, a long headline and several paragraphs compete for the same canvas. Shorter copy or a simpler composition can be more useful than a longer prompt.
4. Choose the right kind of control
| Workflow | Useful when | What to expect |
|---|---|---|
| Natural-language prompt | You want a scene, photograph or visual concept. | A written description for the image model to interpret. |
| Structured layout JSON | You want to describe elements and their intended spatial regions. | Model guidance using the required layout schema, rather than an editable design file. |
| Manual poster editor | Your wording and frame positions must stay fixed. | Separate text and image layers, composed and exported in the browser. |
The official MOSS-Image prompt skill covers both caption and spatial-layout workflows. Our prompt generator uses the caption approach. Our poster editor uses local composition, which is a separate process from asking the model to follow a JSON layout.
5. Review before spending on an image
- Does the prompt still describe the subject you asked for?
- Are the visible words correct, with no invented prices or claims?
- Does the composition leave enough room for the copy?
- Are the style, exclusions and image format consistent?
Edit anything that does not fit. Copying a prompt or opening the image workspace does not submit an image request. The image workspace shows the generation step separately.
A practical next step
Try a restaurant, farmers market or iced latte example in the generator, then adapt the brief to your own subject. For a real product photo you already have, open the poster editor, upload the image and add editable text. Uploads are used as supplied; automatic retouching and background removal are not included.