How to Control the Layout and Aspect Ratio of AI Images

Step 1: Define the destination and distinguish shape from size

Write down the intended use, target ratio, final dimensions and anything that must remain visible. Check the destination’s current specifications before producing a batch.

These dimensions illustrate the ratios; they are not guaranteed outputs from every image model or universal specifications for every channel. For example, 1440 × 2560 is also 9:16. A 16:9 file may be 1920 × 1080 or 3840 × 2160, with different pixel counts.

Calculate the shape as width divided by height. A 1536 × 1024 image is 3:2, not 16:9. Do not stretch it until the file happens to match a required width and height.

Step 2: Turn the layout into a short, testable brief

Specify four things before adding visual style:

• Placement: where the main subject belongs, such as the right third.
• Scale and framing: how much of the subject is visible and how large it should feel.
• Space: where you need quiet background for text or interface elements.
• Constraints: which parts must remain inside the frame and what should not be added.

For a horizontal product image, a useful brief is: product on the right, full handles and base visible, low-detail space on the left for a headline, level camera and generous outer margins. For a tall character portrait, it may be: full figure centered, head and feet inside the frame, extra sky above, and no critical details near the bottom.

Describe empty space as an intended visual quality. “Low-detail continuous background on the left” is often more useful than “empty panel,” which may encourage a visible rectangle. “Leave room for text” also does not require the model to generate the text itself.

Percentages can communicate approximate intent, but they do not create measured guides or pixel-accurate positioning. If a logo must begin at a precise coordinate, place it later as an editable element.

Step 3: Set the ratio in Dreamina and add the appropriate references

Open the Dreamina AI image generator, choose the image model available for your task, and select a supported ratio in the generation controls. Add your composition brief to the prompt so the requested content fits that shape.

If an existing person or product matters, upload its reference and confirm the thumbnail. If you have a rough layout image, explain that it controls placement rather than appearance. Keep the jobs of the references separate: a product photo identifies the item, while a sketch indicates where it should go.

Clear references from previous projects. A leftover portrait, style board or product version can change the output without any obvious error in the written prompt.

For a simple horizontal composition, try this untested starting prompt and adapt it to your subject:

Create a wide landscape product photograph using the uploaded tote as the product reference. Place the full tote in the right third, with both handles and the base completely visible. Keep the left half as a continuous, low-detail warm-gray studio background for editable copy. Use soft light and a grounded contact shadow. Keep generous outer margins. No text, no extra products and no visible rectangular text panel.

Select the matching landscape ratio in the controls. A prompt containing “16:9” does not override unsupported model dimensions or prove the downloaded file has that ratio.

For workflows with several inputs, use the same reference-role method described in combining multiple reference images with AI.

Step 4: Generate and assess the composition before adding final text

Download or inspect the result at full size, then check the intended layout. Ask whether the subject is on the correct side, complete and appropriately scaled. Check the actual negative space: it may be free of objects yet still too textured for readable copy.

The existing reference-led examples in Dreamina show why this is a separate review. A requested ratio can be correct while typography or margins still fail.

In a fictional tote campaign, a vertical image was generated in Dreamina with Seedream 5.0 Pro using a 9:16, 2K setting. An initial readability problem led to a targeted revision that increased text size and contrast. The revised native file was 1440 × 2560, so the ratio was correct. Its small “FICTIONAL DEMO” line still fell inside the bottom 10% area reserved by that demonstration’s layout brief.

The offer and dates in this image are fictional layout content, not a current promotion. The example demonstrates a remaining placement problem, rather than a fully approved layout. The 10% margin was the test’s chosen condition; it is not a universal safe-area rule for every app or device.

The same campaign’s generated horizontal version also had the correct shape, while small supporting text remained difficult to read at a 360-pixel display width after revision. Ratio, readability and safe placement therefore need independent checks.

Step 5: Choose crop, expansion or recomposition

Use the least disruptive operation that can satisfy the brief.

In Dreamina, the AI image expander is an option when the task is to extend surroundings. Upload the approved image, choose the available expansion controls or target framing, describe the added area if the workflow offers a prompt, and generate. Inspect the join, shadows and any subject edges, then download the approved result. This describes the supported editing route, rather than an expansion test performed for this article.

If a whole person is cut off, expanding the canvas may create the missing parts, but it cannot verify their original appearance. A real source view is preferable when those details must be accurate.

For a layout that is almost right, request one change at a time: make the subject smaller, add room on the left, or move the horizon lower. Repeat the important preservation constraints and compare the revised image with the approved reference. A request to retain the subject does not guarantee unchanged pixels.

Step 6: Finish exact layouts on an editable canvas

When the final image needs precise text, product positioning or several sizes, separate the approved artwork from the layout. The following manual route also avoids relying on a paid automatic resize feature.

1
1. Create a new design canvas at the required width and height in a design editor such as Canva.
2
2. Upload the approved artwork or original product image. Fit it proportionally; do not drag one side until the object becomes wider or taller.
3
3. Keep headlines, dates, offers and disclaimers as separate editable text elements. Add a separate background shape if contrast requires one.
4
4. Arrange the product and text for that ratio. Use alignment tools and guides to keep the intended margins.
5
5. Create the next required canvas, copy the editable elements and rearrange them. A tall layout usually needs a different hierarchy from a wide one.
6
6. Review each design at its likely display width, then download and inspect the actual exported file.

An existing manually assembled square master used the original fictional tote photo with separate text and design elements. It was then copied and rearranged for vertical and horizontal delivery. The process retained the source photograph’s rectangular studio background; it did not generate a new scene or redraw the bag.

The tall version placed the offer above the product, with the date and demonstration label below. Its native PNG measured 1080 × 1920. In that example, the text was inside the chosen margins and remained readable at a 360-pixel preview width.

The wide version used a different arrangement: text on the left and the full product on the right. The native PNG measured 1920 × 1080. Supporting text was still relatively small at a 360-pixel preview width, which is a reason to judge the intended display rather than treating export success as universal readability.

These were manual layouts completed in the recorded Canva Free workflow, not one-click AI generation or automatic Magic Resize. Current Canva documentation places Magic Resize on eligible plans and counts its use against the applicable monthly AI limit. If you use that feature, select the required size, create a copy where appropriate, and still review the result.

Other generators: use their actual controls

The same three-part method works elsewhere, but control names and limits vary.

In ChatGPT Images, describe the desired framing and placement and use the available aspect-ratio controls for creation or editing. Its current image guidance supports targeted revisions and explicit spatial language. Check the downloaded shape rather than treating every ratio mentioned in a conversation as a guaranteed native size.

In Midjourney, use the supported aspect-ratio parameter, such as --ar 16:9, alongside the composition prompt. Editing and reference behavior depend on the chosen version and mode. A ratio parameter controls shape, while text and reference instructions still guide placement.

Use only the sizes and ratio controls offered by the selected model in the app. If the required dimensions are unavailable, finish the approved image on a measured design canvas rather than assuming a different model has the same controls.

Step 7: Verify the exported image

Reopen the downloaded file and inspect its properties. Confirm width, height, ratio and format. For a 1080 × 1920 target, receiving a different 9:16 size means the shape is right but the exact delivery dimensions still need finishing.

View the image at its intended display width. Check that the full subject, labels and important text remain readable, with appropriate margins. Preview it in the destination’s crop or interface when possible. Device overlays and thumbnail crops can cover content that looked safe on the design canvas.

Keep the editable master and separate exported versions. Name each by its intended use and dimensions so a later upload does not substitute the wrong crop.

Frequently asked questions

Why does the output have the right ratio but the wrong layout?

The ratio sets the outer shape. It does not determine subject position, negative space or text size. Add explicit composition instructions, use a layout reference where supported, and finish exact elements in an editor.

Can I change a square image into a vertical one without distortion?

Yes. Crop if the important content fits, extend the surroundings, or rearrange separate elements on a vertical canvas. Avoid stretching the whole image to a new shape.

Does “2K” tell me the exact width and height?

No. The resulting dimensions depend on the model, ratio and export workflow. Inspect the downloaded file rather than deriving exact dimensions from the resolution label alone.

Are percentage positions and margins exact AI controls?

Usually they are instructions about approximate composition. Use measured guides and editable layers when placement must be exact.

Start in Dreamina by choosing the ratio and specifying the layout deliberately. Approve the artwork, repair the framing with the appropriate method, and finish exact dimensions and typography on an editable canvas before exporting.