Set the shape and the composition separately
To control an AI image’s layout, choose the canvas ratio in the generator, then describe where the subject and supporting elements belong. A ratio such as 16:9 determines the shape of the image; it does not tell the model to put a product on the right or leave the left side suitable for a headline.
In Dreamina, use the available aspect-ratio setting with a brief that describes subject placement, framing, background, and the area reserved for copy. Seedream 5.0 Pro supports prompt- and reference-guided image creation, making it useful for developing a composition around an existing product or creative direction. Preserve your approved source when exact product detail matters, and judge the downloaded image at its intended display size.
If you already have a finished image, the right method depends on what needs to change. Crop to remove expendable edges, extend to create additional surroundings, or recompose when the subject itself needs a different size or position. For a finished advertisement whose words must remain editable, rebuild the layout around the image rather than regenerating all the lettering.
Choose the canvas before you choose the camera angle
Start with the space the image must occupy. A landscape hero, a portrait poster, and a vertical mobile placement make different demands on the same subject.
These are composition choices, not a claim that every platform requires those ratios. Use the actual placement’s dimensions and overlay rules when preparing a final asset.
Ratio and resolution are separate settings. A 2560 × 1440 file and a 1280 × 720 file are both 16:9, but they contain different amounts of image detail. Conversely, changing a file’s pixel dimensions without maintaining its proportions can stretch the subject. Choose the intended shape first, then select an available output size that suits the placement.
For a single product, avoid starting with a tight close-up if you will later need its entire silhouette in a tall format. A little usable margin around the original often gives you more choices than a frame filled edge to edge.
Write a layout brief the model can act on
A useful brief separates four decisions: what matters most, where it sits, how it is framed, and which space has a job of its own.
Name the visual hierarchy
Say what viewers should notice first. “A perfume bottle, curved glass, leaves, dramatic shadows, large text, and a textured wall” gives six elements similar emphasis. “The perfume bottle is the primary subject; the glass and shadows are quiet background elements” establishes a hierarchy.
For an illustration, the main subject might be a character’s face. For a product advertisement, it may be the package and a short offer. For a thumbnail, it may be an object large enough to recognize beside a compact headline. The placement instruction should support that choice.
Describe position and scale together
“On the right” is incomplete if the product ends up too small to recognize. Specify the part of the frame and the view: a full bottle in the right half, its cap and base visible, with the front label facing the viewer. For a character, say whether the frame must include the head, hands, and feet.
Proportions can help communicate intent, but a prompt asking for a particular percentage is not a measured guarantee. Use the generated composition as a draft you can inspect. If a precise border or grid is mandatory, finish that part in an editor with direct placement controls.
Give empty space a purpose
For copy, describe a relatively even area with low visual detail. “Empty left side” can still produce bright reflections or leaf shadows there. Those elements may look attractive yet make a headline hard to read.
A practical prompt template is:
Create a wide 16:9 product image. Place the complete product in the right half, with its label facing forward. Keep the left third a smooth, evenly lit background suitable for a short dark headline. Confine decorative shadows and reflections to the product side. Do not add text or extra objects.
This is a suggested template, not the exact prompt used for the examples below. Adapt the instructions to the subject and the actual words you intend to add.
Build a new composition in Dreamina
Free accounts in the US receive 120 credits per day; daily allowances vary in other regions. You can start exploring image ideas with Seedream 5.0 Lite and Nano Banana. Open Dreamina, enter AI Image, and choose Lite or Nano Banana for the free starting route. For the Pro examples below, select Seedream 5.0 Pro with the separate model access it requires. Decide whether you need a new image or a reference-led revision before generating.
- 1
- Prepare the reference, if needed. Use an approved image in which the important details are visible. Identify which parts are factual constraints and which are open to creative interpretation. 2
- Choose the target ratio. Use the generator’s ratio setting rather than relying entirely on the words “wide” or “vertical.” Select an available resolution separately. 3
- Write the visual brief. Describe the subject, its position and scale, the camera view, the supporting environment, and the space reserved for other design elements. 4
- Review the first result as a layout. Check what the eye sees first, whether the full subject is visible, and whether the reserved area can actually hold the intended copy. 5
- Revise the failed instruction specifically. If the composition works except for a busy text area, request a simpler background there. If the subject is cropped, ask for the whole subject and more surrounding space. 6
- Download the file and check its actual dimensions. Keep the original reference and the selected output so you can compare changes and return to a sound starting point.
When combining several visual inputs, assign each a role. A product photograph can define the item while a separate layout reference defines arrangement. If those roles are left implicit, a model may borrow the wrong object or composition. The guide to combining multiple reference images develops that choice further.
What three real layouts reveal
The following October 2026 examples use synthetic products and actual Dreamina outputs. They illustrate different layout decisions; they are not a success-rate study.
A wide thumbnail can balance a product and a large headline
Dreamina output: 2560 × 1440 JPEG. The requested headline reads “ONE PHOTO, THREE ADS.”
The image places lettering on the left and the red cup on the right. Broad color fields separate them, and the full headline remains readable in the small 320 × 180 preview. This is a useful composition for a thumbnail that needs to communicate at a glance.
The useful lesson is to plan the amount of copy with the image. A long subtitle or a second product would need a different allocation of space; adding either afterward would compete with the existing hierarchy.
A beautiful background may still fail as copy space
Dreamina output: 2560 × 1440 JPEG, using a fictional perfume reference.
The stone surface, glass, and warm reflections create a coherent scene. The bottle has room on the right. However, the requested left-side copy area contains large shadows and highlights. At a reduced display size, those details still compete with text.
Moving the subject was therefore only part of the task. A suitable revision would simplify that left region while preserving the useful product-side lighting. Alternatively, an editable solid panel could provide a deliberate text surface. Simply making the headline larger would not resolve every contrast problem in the original background.
A tall adaptation can retain the subject without delivering blank space
Dreamina adaptation: 1440 × 2560 JPEG. The complete cap and base remain visible.
The tall version keeps the bottle recognizable and places it in the lower half. It also retains decorative shadows above the product. That is useful for a portrait scene, but it is not the same deliverable as an uncluttered portrait advertising layout.
Before adding copy, decide whether that upper scene should remain part of the visual story or become a simpler information area. Changing orientation successfully does not settle that design decision.
Adapt an existing image without distorting it
Use the smallest intervention that fits the new placement.
Crop when the edges are expendable. Cropping preserves the retained part of the source image. It is effective when the subject already has enough surrounding space and the new frame does not cut off important details. Check the crop around handles, hands, packaging, and the base of a product—not just its center.
Extend when the subject is right but the canvas is too small. Generative expansion can create more background around the existing composition. The new surroundings still need inspection for repeated objects, implausible shadows, and unexpected lettering. Keep the original image available if the extension also changes a region you needed to retain.
Recompose when the visual hierarchy needs to change. Turning a wide product-and-headline arrangement into a tall layout often requires moving or resizing elements. A new generation can explore the scene, while a layered editor can place the approved product and editable copy directly. Choose the latter when preserving exact artwork is more important than inventing a new composition.
To crop an existing image, open it in an editor, fix the crop frame to the target ratio, and move the frame until the important subject fits. Save a separate copy rather than stretching the original into the new shape.
To try a new portrait composition in Dreamina, enter AI Image, choose Seedream 5.0 Pro, and upload the accepted landscape image as the reference. Select 9:16, then request the new arrangement—for example: “Recompose this scene as a 9:16 portrait. Keep the complete bottle, including cap and base, in the lower half; preserve its visible label and material. Build a simple background above it for a headline.” This is a suggested editing instruction, not the submitted prompt for the displayed adaptation. Generate, compare the product and layout with the source, then download the selected file and confirm its dimensions. This route creates a new reference-guided composition; it is not a crop or a promise to preserve the original pixels.
For campaigns with multiple finished sizes, the Black Friday layout adaptation guide explains how to carry an offer between formats. Its job differs from this guide: here the decision is how to create the image’s composition; there it is how to preserve a complete advertisement’s information hierarchy.
Finish at the size your audience will see
Open the export at both full size and a realistic display size. At full size, check the subject’s detail and the joins between original and generated regions. At display size, check recognition, hierarchy, and whether a headline can be read without zooming.
If your final design needs precise words, add those as editable text after you approve the picture. If you want the words rendered within the artwork itself, compare AI image generators for text and verify the actual output character by character.
A controlled AI layout comes from a sequence of concrete choices: select the canvas, establish the main subject, assign the surrounding space a purpose, and revise the part that failed. Dreamina can develop those compositions from prompts and references; your final layout should make the intended subject and message clear in the placement where they will be used.