Skip to content

Improve product photo quality without changing the product

Most weak product photos have one of three problems: they are too small for zoom, grainy because they were shot indoors on a phone, or blocky from being saved too many times. This page shows how to tell which problem you have, which PhotoRescue model fixes it, and how to check the result before it goes on your store.

How it works

  1. 1

    Check the photo at 100 %

    Decide whether the main problem is size, grain or compression blocks.

  2. 2

    Run the matching model

    Enhance for grain and compression, Upscale 4× or 8× for size. Clean first, then enlarge.

  3. 3

    Compare before you publish

    Check labels, seams and edges against the real product before it goes on your store.

Why product photos look worse on the store than on your phone

A phone screen shows a photo far smaller than its real size, so almost everything looks sharp there. A store page does the opposite. It stretches the file to fill a large gallery, and shoppers zoom in to check stitching, labels and surface finish. Every weakness in the file shows at that point.

The causes are easy to spot once you know them. Supplier images are often small, because they were made for an old catalogue. Photos taken in a stockroom or garage carry sensor noise, which the phone tried to hide by smoothing, so fabric and wood look waxy. Files that went through a chat app or a marketplace export have been compressed more than once, which leaves square blocks in plain backgrounds.

Which tool to use for which problem

Open the photo at 100 % zoom and name the main problem. Then pick the tool that matches it:

  • The photo is clean but too small: use Upscale 4×. It makes the photo four times wider and taller. It accepts photos up to 5 MP and gives results up to 80 MP.
  • The photo is tiny, such as an old 400-pixel catalogue image: use Upscale 8×. It accepts photos up to 1.25 MP and uses 2 credits per photo.
  • The photo is the right size but grainy, blotchy or smeared from a dim room, or damaged by a chat app: use Enhance. It keeps the size and removes noise and compression damage.
  • The image is a graphic, a label close-up or a banner with blocky JPEG edges: Enhance removes that compression at the same size too. It was built for photos, so check flat colour areas and small print at 100 % before you publish.

When a photo has two problems

Many supplier photos are both small and damaged. Clean first, then enlarge. An upscaler treats noise and compression blocks as part of the picture, so it enlarges them along with the product. Run Enhance on the original, then run Upscale 4× on the enhanced file.

Each run is a separate job and uses its own credits. On the free plan both results carry a watermark, and the second run would process the first watermark. Test the look with one photo on the free plan, then do the real two-step work on a paid plan.

Why faithful repair matters for products

Some AI tools redraw a product to make it look sharper. On a landscape that may be harmless. On a listing it is a problem: a redrawn knit may not match the real knit, and a sharpened finish may look brushed when the real one is polished. A shopper who receives something that does not match the photo is a likely return.

PhotoRescue's models remove damage and rebuild edges and texture from what is in the pixels. Enhance adds natural fine texture where noise or compression wiped it out, so a sweater looks like knit again rather than plastic. That texture is plausible, not a record of the exact original, so always compare the result with the real item.

Labels need extra care. The models detect text and protect it, so readable text is kept rather than redrawn. Small, dense print can occasionally be missed, and print that is unreadable in the original stays unreadable. Read every label in the result before you publish it. Our guide to faithful and generative upscaling explains the difference in more detail.

How big product images need to be

Each platform sets its own limits, and they change. The table lists only figures we checked on the platform's own help page on the date shown. For other marketplaces, check the seller help centre before you resize anything, and keep your original files.

As a working rule, aim for at least 2000 pixels on the shortest side of a square product photo. A 500-pixel photo reaches 2000 pixels with one Upscale 4× run. If the result is larger than the platform accepts, reduce it in any editor. Reducing a sharp image keeps it sharp.

A simple workflow for a whole catalogue

Sort the images first: one folder for photos that only need enlarging, one for photos that need cleaning. Test three typical photos on the free plan, which gives you 3 free credits a day with no sign-up and no card, and compare them with the originals.

Free results carry a watermark, so they are for testing, not for publishing. Paid plans remove the watermark, accept uploads up to 20 MB and 30 MP, and process up to 50 photos in one batch. Pro includes 200 credits a month for $9.99 and Pro+ includes 500 for $19.99. One photo with one model uses 1 credit; Upscale 8× uses 2.

When to reshoot instead

No tool recovers a product that was out of focus, blurred by movement, or so small that the details are only a few pixels across. In those cases a quick reshoot next to a window beats any repair.

PhotoRescue also does not remove backgrounds, retouch products or change colours. Do those steps in your usual editor, before or after the repair.

Product image limits from the platform's own help page (checked 2026-10-09)

Product image limits from the platform's own help page (checked 2026-10-09)
PlatformWhat the help page says
ShopifyProduct and collection images up to 5000 x 5000 px or 25 megapixels, file size under 20 MB
Shopify2048 x 2048 px usually displays best for square product images

What it won't do

  • It does not recover detail that is not in the photo, such as unreadable label text or a pattern lost to blur.
  • Texture added by Enhance is plausible, not a record of the exact original; compare the result with the real item.
  • It does not remove backgrounds, retouch products or change colours.
  • Strong motion blur and badly out-of-focus shots need a reshoot.
  • Free results are watermarked, and free uploads are limited to 3 MB and 8 MP.

Questions

Can AI improve product photos without changing the product?

It can if it repairs rather than redraws. PhotoRescue removes noise, compression and resize blur and rebuilds edges from what is in the pixels. Always compare the result with the real item before publishing.

What size should product photos be?

It depends on the platform. Shopify's help centre says 2048 x 2048 pixels usually displays best for square images and accepts up to 5000 x 5000 or 25 megapixels. Check your platform's own help page.

How do I make a small supplier image bigger without blur?

Use Upscale 4× on the cleanest copy you can get, or Upscale 8× if it is under about 1.25 MP and you need a large result. If it is grainy or blocky, run Enhance first and upscale the enhanced file.

Will it change the text on my labels?

The models detect text and protect it, so readable text is kept rather than redrawn. Very small or dense print can occasionally be missed, and unreadable print stays unreadable. Read every label in the result before you publish.

Is the free plan enough for my store?

The free plan is for testing: 3 free credits a day, no sign-up and no card, with a watermark on results. For images you publish, use Pro or Pro+, which have no watermark.

Can I process my whole catalogue at once?

Paid plans process batches of up to 50 photos. Pro includes 200 credits a month and Pro+ 500. Upscale 8× uses 2 credits per photo; every other model uses 1.

Will it remove the background?

No. PhotoRescue only repairs image quality. Do background work in a dedicated editor.

Processing photos in bulk or inside your own app? Use the image enhancement and upscaling API