Compress Product Photos Safely With Squoosh

Large product photographs are useful while you are editing, but they can make a storefront slower and are often awkward to upload. The obvious fix is compression. The risk is that an image can look acceptable at fit-to-screen size while a logo, hallmark, chain link or ingredient line has already started to break apart.

Squoosh is a free browser-based image optimiser that lets you compare the original and compressed result side by side. Its official site says processing happens locally on your device, so the photograph itself is not sent to a server. That makes it a practical place to build an export recipe before applying similar settings elsewhere.

This workflow is for delivery copies, not masters. Keep the original camera file, layered edit and final high-quality export. Compression should create a smaller derivative for a specific channel, never replace the files you would need for a later crop, colour correction or print job.

1. Start with an approved master

Finish the product edit before opening Squoosh. Confirm colour, dust cleanup, background, crop and any exact text or logo work first. Repeatedly saving a lossy JPEG can compound damage, so always compress from the approved high-quality master rather than from a file that has already passed through several uploads and downloads.

Use the final pixel dimensions required by the destination. If your image is already the correct size, leave Squoosh’s resize control off. If it is much larger than necessary, resize once as part of this export. Do not enlarge a small image here. More pixels do not restore missing texture or readable type.

2. Decide whether the file needs JPEG, WebP or AVIF

For a photographic product image going to a marketplace or a system with uncertain format support, MozJPEG is the conservative starting point. It produces a standard JPEG and gives you an adjustable quality trade-off. JPEG does not preserve transparency, so a cutout that needs an alpha channel should remain PNG or use a format the destination explicitly accepts.

For a storefront you control, compare WebP and AVIF as well. Squoosh supports MozJPEG, WebP and AVIF alongside other codecs. Google describes WebP as supporting both lossy and lossless compression, including transparency. Modern formats can create smaller files, but a smaller number is not useful if your marketplace, theme, feed or advertising tool rejects the upload or silently converts it.

Check the destination’s current rules before committing a catalogue. If the platform already creates responsive WebP or AVIF versions from an uploaded JPEG, supplying your own modern-format file may add work without changing delivery.

3. Use the split preview as a defect detector

Load one representative image at squoosh.app. Keep the original on one side and choose the export codec on the other. Move the comparison line across the product rather than judging the whole frame from a distance.

Zoom to 100 percent and inspect the smallest truthful details: engraved letters, label text, gemstone facets, fine chain links, stitching, patterned fabric and hard edges against the background. Lossy compression usually becomes objectionable in these areas before it is obvious in a smooth wall or soft shadow.

Start at a generous quality setting, then lower it gradually. Stop when you first see ringing around a dark logo, blocky colour transitions, smudged texture or colour bleeding along a saturated edge. Move the quality back up until the defect disappears. This gives you an image-specific threshold instead of an arbitrary quality number copied from another seller.

4. Judge the file at two viewing sizes

A listing image has two jobs. It must look clean as a small search-result thumbnail, and it must survive a customer zooming into the product. Check both. At thumbnail size, watch for thin chains or prongs disappearing. At 100 percent, look for false halos and damaged lettering.

Do not optimise to a file-size target alone. A 150 KB image of a plain ceramic cup may look excellent, while a 150 KB image containing jewellery, textured packaging and fine print may not. The amount and type of detail determine how hard the codec has to work.

Also compare the compressed colour with the master. Reflective metals and subtle stones can show banding in gradients or shifts in small coloured highlights. If the output changes the product’s apparent finish, raise the quality or use a less aggressive format setting.

5. Keep transparency and white backgrounds separate

A transparent cutout and a white-background listing image are different assets. If you convert a transparent PNG to JPEG, the transparency must be replaced with a solid background. Confirm that it becomes the intended white, not black or a grey fringe.

For a master cutout, lossless PNG is often worth the larger size because it preserves the edge for future scenes. Create a separate flattened JPEG or WebP for the storefront. This prevents one highly compressed edge from being reused in every lifestyle image and colour variant.

6. Test one hard SKU before building a preset

Choose the most difficult item in the range, not the easiest. For jewellery, that might be a fine chain with a small hallmark. For cosmetics, use a reflective bottle with fine label copy. For clothing, choose dense fabric texture. Find settings that pass on this product, then test them on several ordinary SKUs.

Record the codec, quality setting, output dimensions and intended channel in a simple export note. Name files so the master and delivery version cannot be confused. A pattern such as SKU_channel_width_format is more useful than final-final-small.jpg.

Squoosh is strongest as a visual comparison tool for individual images. For a large catalogue, use it to establish and verify a recipe, then apply that recipe with a batch-capable exporter. Spot-check the batch because two tools can interpret similarly named quality settings differently.

7. Run an upload test

Before processing hundreds of images, upload three files: a normal product, the hardest fine-detail product and an image with transparency if the channel allows it. View them on the live product page or a hidden test listing.

Check desktop and mobile, open any zoom view, and download the served image if possible. A commerce platform may resize or recompress the upload again. If the live version is visibly worse than the local Squoosh export, increase the source quality slightly or investigate the platform’s image pipeline before changing every file.

The useful result is not the smallest possible image. It is the smallest delivery file that preserves the evidence a customer needs. For jewellery, that evidence can be a clasp, setting or engraving. For packaged goods, it can be exact label text. Once those details start to blur, the file has crossed the line from optimisation into product misrepresentation.

Sources

Squoosh official web app, reviewed 10 October 2026: https://squoosh.app/

GoogleChromeLabs, Squoosh source and privacy notes, reviewed 10 October 2026: https://github.com/GoogleChromeLabs/squoosh

web.dev, “Announcing Squoosh v2,” reviewed 10 October 2026: https://web.dev/blog/squoosh-v2

Google for Developers, “An image format for the Web,” reviewed 10 October 2026: https://developers.google.com/speed/webp