Image generators default to square because square is the shape that costs a model the least trouble. Nobody sells square posters in any volume, because nobody buys square frames in any volume. So every AI poster range starts with a shape mismatch that has to be resolved somewhere, and where you resolve it decides how the whole range looks.
There are three ways out — crop, extend, or mount — and only one of them survives being repeated fifty times.
Cropping is a composition edit
Take a 1:1 image to 2:3 portrait and you are not resizing anything. You are keeping the full height and cutting a third of the width away.
That third is where a square composition does its balancing. Square framing pushes the subject off centre and props it up with negative space on the other side; take the sides off and the subject sits dead centre in a tall slot, with whatever it was in conversation with now outside the frame. Broad landscapes tend to survive. A lighthouse with a bank of cloud balancing it on the left does not — you get a lighthouse and nothing else, and the image goes from composed to snapped.
The tell is that cropped ranges look inconsistent even when every generation used the same style reference, because the crop lands differently on every subject. You made fifty independent composition decisions and called it an export step.
Extending drifts
Generative fill, Midjourney's zoom out, and outpainting in general will all give you the extra height. Each one invents the content that fills it.
For one poster, fine. For a range it is precisely backwards: the invented region differs on every image, so the part of the collection that should be most uniform is the part being improvised. You also lose control of where the subject sits, because you do not know in advance whether the model will add two centimetres of sky or twelve.
There is a resolution cost as well. Extended regions come back at whatever the model gives you, often softer than the original, and they land at the top and bottom of the sheet where the eye enters and leaves the image.
Mounting is what the market actually does
Look at the square-artwork travel posters that sell well and the structure is always the same. The image sits inside the sheet at close to full width, with matching margins at the top and sides, and a deep band of flat paper colour at the bottom carrying the type.
Nothing is cropped, nothing is invented, and the frame is identical on every poster in the range. The image can change completely and the layout does not move a millimetre, which is what makes twelve prints read as a collection rather than as twelve prints.
It also solves the typography problem for free. You have a permanent, dedicated, image-free area for the place name — no text sitting on artwork, no hunting for a quiet patch of sky that happens to be somewhere different on every render.
The band is deeper than you think, and that is geometry
Here is the part people fight and should not.
Put a square image at full width on an A4 sheet with a 20mm side margin. The image is 170mm wide, so it is 170mm tall. Match the top margin to the sides at 20mm and the bottom gets what is left: 297 − 20 − 170 = 107mm.
That is a band more than five times the depth of the top margin, and about 36 per cent of the sheet. On screen it looks like a mistake. Hung at eye level it looks correct, because the eye reads the optical centre of a frame as sitting slightly above the geometric one — the same reason picture framers cut mounts with a deeper bottom rail.
The arithmetic is unavoidable, incidentally. On A4, with the top margin equal to the sides, the bottom margin always works out to 87mm plus the side margin, whatever width you choose for the image. There is no arrangement of a square on a 1:1.414 sheet where three margins match and the fourth is only slightly deeper. Either the band is deep, or the top margin has to grow far beyond the sides, which looks worse. So design the band on purpose and give it a job.
The band changes size with the ratio, which is the real problem
Same square image, same margin logic, four common sheets. What is left over vertically, as a share of sheet height:
- 4:5 (40×50cm) — 20 per cent before margins, a band near 26 per cent after
- 3:4 (30×40cm, 18×24in) — 25 per cent, band near 31 per cent
- A-series (1:1.414) — 29 per cent, band near 36 per cent
- 2:3 (50×75cm, 24×36in) — 33 per cent, band near 39 per cent
The band on a 2:3 sheet is half again as deep, proportionally, as the band on a 4:5 one. Your type therefore sits in a different place relative to the image, at a different relative size, with a different amount of air around it.
This is why a single file cannot serve both the European and the American markets, and it has nothing to do with resolution. A-series is 1:1.414, 3:4 is 1:1.333 — about six per cent apart in height. Fit an A-ratio file to the width of a 30×40cm sheet and you have 24mm of unplanned paper to dispose of. Fill instead and the trim takes a bite out of the margin you carefully set. Scale it non-uniformly to force the fit and everything circular in the artwork becomes six per cent of an ellipse, which is invisible on a mountain and glaring on a moon, a wheel or a dome.
The right mental model is that you do not have one poster file. You have one square artwork and a layout rule, re-run per ratio. Fix the artwork, recompute the margins for each sheet, export each one. That is a template problem rather than an export problem, and it is the sort of thing worth building once — a batch layout tool such as PSDmate exists because doing this by hand across four ratios and forty places is a weekend nobody enjoys twice.
What margins actually do at print size
Three things, and only one of them is aesthetic.
They give the frame something to eat. Almost every frame overlaps the sheet edge by a few millimetres to hold the print in the rebate. Artwork with a 5mm margin loses most of that margin behind the moulding, and any type near the edge goes with it.
They decide whether the buyer can mount it. Plenty of buyers put a window mount around a print. Your printed margin then sits inside their mount board, and a 40mm printed margin plus a 50mm mount reads as an accident rather than a design. Deep printed margins and mounting are alternatives, not a stack — worth one line in the listing description saying which you designed for.
They scale with the sheet, which is fine, unlike type. A 20mm margin on A4 is proportionally identical to a 28mm margin on A3, so a proportional layout holds across the whole A-series without thought. The type inside the band does not behave so kindly, and that is a separate problem needing a separate solution.
When full bleed is the better answer
If the artwork genuinely has nothing to say at the edges — an abstract, a texture, a colour field — printing it edge to edge on a square sheet is a legitimate route, and 30×30cm and 50×50cm frames do exist.
The trade-off is honest and unglamorous. The square frame market is a fraction of the size of the portrait one, and you give up the type band that keeps a range coherent. It suits a particular kind of shop and fails badly for a place-name poster range.
The order that works
Decide the ratios you will sell before you generate anything, because that decision sets the layout, the layout sets where the type goes, and the type is what makes the range look designed.
Then generate square, mount on a fixed layout per ratio, and leave the artwork alone. The image is allowed to be different on every poster. Nothing else is.
Frequently asked questions
Can I just crop a square Midjourney image to 2:3?
You can, but you are cutting a third of the width away, and the width is where a square composition keeps its balance. Square framing puts the subject slightly off centre with negative space holding the other side, and removing the sides collapses that relationship into something centred and static. On broad landscapes it usually survives; on anything with an asymmetric subject it rarely does. Cropping is a composition edit, so judge the result as a new composition rather than assuming a resize happened.
Should I use generative fill to extend a square image to a taller ratio?
For a single poster it is a reasonable option, and for a range it is the wrong one. Outpainting invents new content at the top or bottom of every image, and that invented content differs each time, so the thing your collection is meant to share is the exact thing that drifts. You also get an unpredictable amount of extra sky or foreground, which moves the subject around the frame from poster to poster. Mounting the square on a fixed layout gives you the same frame every time with nothing invented.
How deep should the bottom margin be on a square-on-portrait poster?
Deeper than you expect, because the geometry decides most of it for you. A square image sitting at full width on an A4 sheet leaves about 87mm of spare height, and if the top and side margins are equal that entire remainder lands at the bottom. The result is a band roughly five times the depth of the top margin, which looks wrong on screen and correct on a wall, because the eye reads the optical centre of a frame as slightly above the geometric one. Design the band deliberately and put your type in it rather than trying to even the margins out.
Why can one file not serve both A4 and 18x24 inches?
Because the two sheets are different shapes, not different sizes. The A-series is 1:1.414 and 3:4 is 1:1.333, a difference of about six per cent in height, which is far too large to absorb and far too small to notice until it prints. Fit an A-ratio file to the width of a 3:4 sheet and you have a strip of unplanned paper at one end; fill instead and the trim eats into the margin you set. The fix is to rebuild the layout at each ratio around the same square artwork, which costs nothing once the layout is a template.
Which print ratio suits square artwork best?
4:5 is the friendliest and 2:3 is the harshest. With the artwork placed at full sheet width, 4:5 leaves about a fifth of the sheet height spare, 3:4 leaves a quarter, the A-series a little under a third, and 2:3 a full third. That spare height becomes your bottom band, so the same square image with the same margins produces a band occupying roughly a quarter of a 4:5 sheet and close to two fifths of a 2:3 one. If you sell a single ratio and want the image to dominate, 4:5 is the easiest to make look intentional.