Temu and Shein marketplace listing images: what sellers need in 2026

First, check whether you can actually sell there

Before you spend an afternoon on images, check whether you can actually sell there, because the two platforms are not equally open. Temu runs a local-seller programme in several markets for businesses that hold stock locally, alongside its managed model, with eligibility and fees depending on your country and seller type. Shein Marketplace, by contrast, is a programme for approved brands and suppliers, not an open self-serve signup; final eligibility is confirmed inside the Shein Seller Portal. If you are not approved on Shein, this image set is still useful for your own store, but do not build a plan that assumes a Shein storefront you do not have.

Second, do not list with photos copied from another seller, from Shein itself, or from a supplier catalogue you have no right to use. Buy or make one sample, photograph that sample, and generate scenes around your own product. If a supplier offers images, ask in writing whether you may use them on Temu and Shein, keep the answer, and still check colour and size against your sample.

The specs, side by side, with the disagreements labelled

Both Seller Centers sit behind a login, so the numbers below come from integration and seller guides checked in late 2026, not from a public rulebook. Where guides disagree, this article says so and gives the export that satisfies the stricter version.

Read the file-size row twice. A 2,000-pixel PNG can pass every pixel rule and still fail a 3 MB cap, and a heavily compressed JPG can pass the cap while looking blotchy at zoom. Export as JPG, check the actual file size, and keep the product filling most of the frame. A small product floating in a large white square reads as nothing at all on a phone. If a master is larger than the slot needs, a free image resizer can shrink it to the listing size before you export, so the file-size check starts from the right pixels.

The eight-image set, in the order shoppers ask questions

Shoppers ask the same questions in the same order on both platforms: what is it, what colour is it really, how big is it, what is it made of, what comes in the box, and what does it look like in use. Build one set that answers them in that order, then crop it twice, square and portrait. For apparel, the portrait set leads on Shein; for everything else, square leads.

Image one is the main image: the real product, centred, fully visible, on pure white, no text, no watermark, no price stamp, no collage, no props crowding the frame. Image two is a second angle of the same item, still clean. Image three is a close-up of the detail that justifies the price: stitching, texture, clasp, print, or material edge. Image four is scale: the product beside a common object, or on a model, with no invented size. Image five is the size chart for apparel, or the dimensions graphic for a non-apparel product, in centimetres and inches. Image six is what is in the box, laid out flat, every included part visible. Images seven and eight are lifestyle: the product in a real-looking room or worn outdoors, with the product itself unchanged. Keep the product identical across all eight. Changing its colour, shape, pattern or number of parts between frames is the fastest way to a return, because the buyer will compare the parcel with the prettiest frame, not the most accurate one.

For apparel on Shein, add an on-model photo showing fit on a real body shape, with the size worn stated in the listing text, not baked into the image, and a colour swatch where the category asks for it; guides list an 80 by 80 pixel swatch, and it must match the garment. For Temu, keep titles short and factual: a title promising a feature the images do not show reads as misleading even if every photo is compliant.

Where AI helps, and where it must stop

AI is excellent at the jobs around the product: removing a background, placing your photographed product in a clean room scene, generating a second lifestyle context, or cleaning a distracting background from a real photo. It is dangerous at the job of inventing the product: changing its colour to a more flattering shade, smoothing a fabric texture into something the parcel will not match, adding a pattern, a pocket, a strap or a finish that does not exist, or generating a model wearing a garment in a size and fit the real garment does not achieve. On both platforms the rule is the same even when the wording differs: the image must show what the buyer will receive. An AI background that leaves the product untouched passes that test. An AI edit that improves the product fails it.

Apply the same test to people. A generated model can be a mood or scale reference on your own store, but where fit decides the purchase, a real on-model photo of the real garment, in the stated size, reduces size returns. Never present a generated person as a real customer, reviewer, or proof of fit. Our marketplace disclosure guide covers where a label or a description line is expected, and the background removal guide covers the safest AI edit of all: a clean cutout of your own photo.

Finally, run the pre-upload check on the live upload screen, not on a guide: thumbnail readability, file sizes, consistent ratio, and a size chart legible on a phone. Then duplicate the set for the next SKU. In Movapix an eight-image set, even with rejected variants, costs well under a dollar in generation; the real cost is your time checking colour and scale against the sample. Spend it on images one, three and five.