How to optimize apparel images for marketplaces from one master set

A clothing photo cropped for one marketplace often breaks on the next. Walmart's content policy puts men's, women's, kids' and baby apparel in a 3:4 portrait format. TikTok Shop asks for square images. eBay shows the main photo at 96 pixels wide in search results unless the seller pays for Gallery Plus. One export folder cannot satisfy all three, and stretching a square into a portrait frame makes a sweater look wider than the one that ships.

This guide covers the part most marketplace image advice skips, which is producing one master set and deriving each channel's files from it. What belongs in the gallery is in our guide to the essential images for clothing PDPs. Amazon, Etsy and Shopify each have their own post, linked below, so their rules are not repeated here.

The short answer

Produce every apparel image as a 3:4 portrait master, at least 2400 by 3200 pixels, with the whole garment inside the middle 75% of the frame height. From that master, export 3:4 files for Walmart's apparel categories, square crops for TikTok Shop and eBay, and white-padded squares for hero shots on pure white. Check every hero at 96 pixels wide before upload. Give each color variant its own main image, and name files by SKU, color, angle and ratio so the right file reaches the right channel.

Marketplaces disagree on shape before anything else

Background color and image count get most of the attention in marketplace guides. Aspect ratio causes more rework, because it decides whether the garment survives the crop at all.

Walmart is the clearest case. Its product content policy lists 1:1 square files for most categories and a separate 3:4 portrait specification for men's, women's, kids' and baby apparel. The portrait rule dates to 2018, when Walmart told fashion sellers to replace square and landscape images and said stretched resizes were not acceptable. Several 2026 vendor guides still describe Walmart as square across every category. For apparel, that is wrong. The primary image still needs a pure white background (RGB 255, 255, 255).

TikTok Shop goes the other way. AfterShip's summary of its listing rules, updated in May 2026, gives a minimum of 600 by 600 pixels in square format, a front view on pure white for the main image, and up to nine images per listing. The same summary says every image has to show the product the customer receives.

eBay is the loosest of the three. Its help pages require at least one photo, allow up to 24, set a floor of 500 pixels on the longest side and recommend about 1600. They do not fix an aspect ratio, so the constraint that matters on eBay is the size of the search thumbnail.

For the other three big channels, see Amazon apparel image requirements, Etsy product photo requirements and Shopify clothing image requirements. Specs on every marketplace change by category and region, so confirm them in the seller dashboard before a large upload.

Make one portrait master and derive every export

A 3:4 master is the safest starting shape for clothing. Most garments are taller than they are wide, and a portrait frame keeps a full-length dress large in the image instead of floating in white space. Square files come from it by cropping or padding, and neither needs upscaling if the master is big enough.

The arithmetic is simple. A 2400 by 3200 pixel master gives a 2400 pixel square when you crop 400 pixels from the top and 400 from the bottom. That square is above eBay's 1600 pixel recommendation and four times TikTok Shop's floor. The crop removes 12.5% of the height at each end, so anything outside the middle 75% of the frame is lost. Frame the master with the collar and the hem inside that band and every square export keeps the whole garment.

Shooting or generating at that size costs nothing extra on most setups, and it is far cheaper than finding out after a Walmart upload that every image needs replacing.

Crop the on-model shots, pad the white ones

Padding works the other way. Add 400 pixels of white to each side of the same master and you get a 3200 pixel square with the garment untouched. This only works when the background is pure white to the edges. A light grey backdrop or a lifestyle scene shows the join as a visible band, so those images get cropped instead.

We'd pad the hero and any ghost mannequin or flat lay on white, crop the on-model and detail shots, and never stretch anything. A stretched knit looks a size wider than the real one, which is a fit complaint waiting to happen.

If the source has an uneven or off-white backdrop, fix that before padding. Smart Retouch cleans backdrops and lighting in one pass, and a padded square only looks right once the white is true white.

Check the hero at 96 pixels

eBay shows the main photo at a standard 96 pixels in search results, and at up to 400 pixels on hover only with Gallery Plus. Search grids on other marketplaces are not much bigger on a phone. Shrink your hero to 96 pixels wide and put it next to two competitors' listings. You should still be able to tell the category, the color and the neckline or leg shape.

Dark garments fail this test most often. A black knit on white loses its outline and reads as a dark shape with no neckline. Thin straps and fine prints disappear too.

The other common failure is scale. A garment that fills half the frame turns into a speck at 96 pixels, and Walmart's image guidance tells sellers to avoid excess background space around the item for the same reason. Fill the frame for search, then check that the tighter crop still leaves margin around sleeves and hems at full size.

Variants and file names across channels

Walmart's item image guidance asks for a main image showing the current variant, especially when the variant is a color. It also says that if one item in a variant group has front, back and side views, every item in the group should have them. That is a sound rule on any marketplace. A shopper who picks the sage colorway and sees a black hero has no reason to trust the size chart either.

Color accuracy and the variant gallery itself are covered in the flat lay, ghost mannequin and on-model guide and the essential images post. The marketplace problem is logistics. A drop of 20 SKUs in four colors with five images each is 400 images, and three export ratios turn that into 1,200 files.

Name files so each one can be routed without opening it. We use SKU, color, angle and ratio in that order, so a back view of a sage knit becomes KN204_sage_back_3x4.jpg for Walmart and KN204_sage_back_1x1.jpg for TikTok Shop and eBay. Keep the masters untouched and treat every export as disposable. When a marketplace changes its spec, you re-export from the masters instead of reshooting.

AI-generated sets and the real-product rule

Walmart and TikTok Shop both tie listing images to the product the buyer receives. A generated model shot of your real garment can meet that bar. A generated image that adds a pocket, moves a seam or shifts the dye shade does not, however good it looks in a thumbnail.

Check generated images at full size against the source garment. Look at logos, print alignment, button count and color. Be stricter with back views. Our multi-angle tool makes front, back, side and three-quarter views from one image on the same model, but a back view made from a front photo guesses at details the source does not show. If the back has a yoke, a print or a zip, photograph it.

If you sell into the EU, generated model images may also need a disclosure. The EU AI Act guide for product photos covers which images and from when.

Before you upload to each marketplace

• The master is 3:4 and at least 2400 by 3200 pixels, with the whole garment inside the middle 75% of the height.
• Walmart apparel files are 3:4 portrait, and no image has been stretched to fit.
• TikTok Shop and eBay files are square crops or white-padded squares at 1600 pixels or more.
• The hero still shows category, color and silhouette at 96 pixels wide.
• Every color variant has its own main image, and every variant in a group has the same set of angles.
• File names carry SKU, color, angle and ratio.
• Generated images have been checked against the real garment at full size, back views first.