The resize is the cheap part. Build one master, lock the brand elements, extend instead of crop, and every placement and market becomes a derivation instead of a redesign.
One approved ad rarely stays one file. Five placements and four languages turn it into 20 exports, each with its own layout fixes, its own copy check, and its own round of approvals. The resize is the cheap part. The re-layout and the re-approval are what eat the week.
The fix is a workflow, not a faster export button. Build one master that is designed to be adapted, derive every placement and market from it, and change only what each version needs. Here is how to do that, including the ad sizes worth building for in 2026 and where AI helps without taking the layout decisions away from you.
Most ad masters are built for a single placement, usually the square feed post, and everything else is an afterthought. When the Story version is needed, someone crops, discovers the headline now sits under the profile bar, moves it, and the logo drifts.
The German version arrives, the headline runs 30% longer, and the CTA button no longer fits. Each fix is small. Across a campaign, they add up to a second production cycle nobody budgeted for.
The underlying problem is that the master carries no rules. It is a flat picture, so every adaptation is a redesign. Give the master a structure, and adaptation becomes a sequence of controlled changes instead.
Before you resize anything, decide what in the ad is fixed and what is allowed to move. That one decision does most of the work.
Logo, headline, CTA, and palette stay identical across every version. Treat them as a locked group: the same lockup, the same type hierarchy, the same color values. If your team works with a brand kit in Recraft Studio, these are already defined, which removes the eyeballing of hex codes later. Everything that is not in the locked group is adaptable by default.
Cropping from 1:1 to 9:16 throws away a third of the image. Extending keeps the whole composition and adds canvas. That only works if the background can grow: keep the subject away from the edges, avoid baked-in text or patterns that would need to continue perfectly, and prefer scenes with a natural horizon, gradient, or texture.
A product on a studio sweep extends cleanly. A product in front of a shop sign with legible lettering does not.
Never flatten the headline or CTA into the image. Live text is what lets you swap the string for a new market without touching the visual, and it is what keeps typography identical across 20 versions. If the master arrives flattened, rebuild the text layer before you do anything else. It takes ten minutes and saves hours.
You do not need every size a platform supports. Five formats cover the large majority of paid social and display placements, and most networks accept at least three of them.
Build the master at 4:5 or 1:1 with generous background, and the other formats are extensions rather than crops. Platforms update their specs regularly, so confirm the current numbers in Meta, LinkedIn, and Google's own documentation before a large flight.
Recraft Studio gives you two routes from one master to every aspect ratio. Both keep the composition and change only the canvas, so each format is a new composition rather than a crop. Which one you pick depends on how much of the original you need to keep pixel for pixel.
Outpainting extends the image beyond its original borders and leaves the original pixels untouched, which makes it the safer route when the product shot or the model has already been approved.
The second route treats the master as a reference rather than a fixed image. Select Edit image, attach the master as a reference, pick an editing model such as Recraft V4.1, Nano Banana 2, or Seedream 5.0 Pro, set the output aspect ratio to the target format, and describe what you want: "the same ad composition at 9:16, keep the product, logo placement, and palette, extend the backdrop vertically." The model recomposes the scene for the new shape instead of only adding canvas.
This route is faster when the layout has to change more than the edges, for example moving from a wide 16:9 banner to a tall Story where the product should sit higher in the frame. The trade-off is that it regenerates rather than extends, so check the product, type, and logo against the original before you treat the output as approved. Outpainting for fidelity, reference editing for flexibility.
Whichever route you take, check each format on a phone before you export. Text that reads at 1:1 on a desktop monitor can disappear at 4:5 on a six-inch screen, and Story overlays hide more than people expect. Neither route is magic: busy backgrounds with repeated objects or legible text can produce visible seams or drift, which is the reason the master's background should be designed to extend in the first place.
Localization is where single-file masters fail hardest, because translated copy changes length, direction, and sometimes meaning. The workflow is the same as for formats: change only what the market needs.
French, German, Spanish, and Italian regularly run 15% to 30% longer than English, and single German compound words can refuse to wrap at all. Arabic and Hebrew read right to left, which mirrors the reading order of the entire layout. Design the English master with that slack built in: a headline that fills 90% of its box in English will not survive translation. Aim for 60% to 70% so the longest market still fits without a font-size change.
Because the copy is live text, each market is a text swap on a duplicate of the format, with the same font, weight, size, and color. Translation itself happens in your own process, with a native speaker or your localization partner. The design rule is that no market gets its own typography. If a language needs a different script, choose a companion typeface for it once and apply it everywhere.
Sometimes the copy is not the only thing that has to change. A product variant sold in one market, a seasonal scene, or a prop that reads differently in another culture. Natural language editing in Recraft Studio changes only what you describe and leaves the rest of the image intact, so you can swap a winter coat for a linen shirt in the Spanish version without regenerating the ad. For smaller fixes, inpainting handles a single object or region.
Several tools now do the full chain in one step. Luma Variants, launched in October 2026, takes one approved static ad and returns five formats plus translated copy with the logo, headline, and CTA preserved. Ideogram's Ad Resizer and Scenario's multi-format generator do the resizing part. For high-volume campaigns with simple layouts and low creative variance, that is the right call: speed wins, and the output is good enough for most placements.
The trade-off is that the tool makes the layout decisions. Where the subject lands in the Story, how the headline breaks in German, and whether the background extension matches the art direction are all chosen for you, and reviewing 20 auto-generated files is its own job.
Recraft Studio sits on the other side of that line. You keep the decision at every step, and you still skip the rework, because the master, the locked group, and the live text carry the structure for you. Use the automated route when the ad is simple and the volume is high. Use the Recraft Studio routes when the creative matters.
Treat the master as a system rather than a picture, and every new placement or market becomes a derivation instead of a redesign. The first campaign takes slightly longer to set up. Every campaign after it takes a fraction of the time. If you want to try it on your own ad, open it in Recraft Studio, expand it to 9:16, and see how much of the layout survives.
.svg)
.webp)
