Why Product Photos Look Soft After Resizing (2026 Guide): Resampling, Output Sharpening and Marketplace Recompression

You shot a crisp product photo, resized it for the listing, uploaded it, and now it looks soft. The edges are mushy, the fabric texture has turned into a blur, and the small print on the label is unreadable. Nothing went wrong with the camera. The softness crept in during resizing and upload, and every stage of that pipeline is fixable once you know what it does to your pixels.

This guide walks through the three places product photos lose sharpness after the shoot: the resampling algorithm that shrinks the image, the missing sharpening pass that should follow it, and the second round of compression the marketplace applies after you upload. Each one has a practical fix.

Why shrinking a photo makes it soft

A 24-megapixel camera file is roughly 6000 pixels wide. A Shopee thumbnail is 1024 pixels. To get from one to the other, the software has to throw away about 97 percent of the pixels and work out what colour each surviving pixel should be. The maths used for that decision is called resampling, and the choice of algorithm has a bigger effect on perceived sharpness than most sellers realise.

The three common resampling methods

• Bilinear looks at the 4 nearest source pixels and averages them. It is fast and smooth, but it discards fine detail and produces visibly softer results on large reductions. Many quick online tools default to it.
• Bicubic samples 16 surrounding pixels along a cubic curve. It is the long-standing default in Photoshop and holds up well for moderate reductions of up to around 50 percent.
• Lanczos uses a windowed sinc filter that samples a wider area still. It preserves edges and texture noticeably better when you shrink aggressively, which is exactly what happens when a camera file becomes a 1000-pixel listing image. It is the default kernel in the libvips and sharp libraries that power many modern image pipelines.

In one measured test on a photo-like image reduced by half, Lanczos kept about 87 percent of the original edge sharpness while bilinear kept about 71 percent. That gap is the difference between a product shot that looks professional and one that looks like a screenshot.

Aliasing and moire on patterned products

Poor downscaling does more than soften. When a tool skips pixels instead of averaging them properly, fine repeating patterns such as knit fabric, mesh, speaker grilles and woven straps develop shimmering moire bands that were never in the original. If your textile or electronics photos show rainbow ripples or jagged lines after resizing, the cause is almost always a cheap resampler. Switching to Lanczos or a proper area-averaging method fixes it at source.

The sharpening pass most sellers skip

Even the best resampler blurs slightly, because averaging pixels is inherently a smoothing operation. Professional retouchers compensate with a dedicated output sharpening step applied after the resize, never before it. Sharpening the full-size file and then shrinking it wastes the effort, since the resampler averages most of the sharpening away.

What output sharpening actually does

Output sharpening is a light unsharp mask tuned to the final pixel size. It raises contrast along edges by a small amount so that the image reads as crisp on a screen at 100 percent. The key word is light. Over-sharpening creates bright halos around product edges and crunchy noise in smooth backgrounds, and it also inflates the JPEG file size because the encoder has to describe all that extra edge detail.

Sensible starting points

• Lightroom and Camera Raw: in the export dialogue set Sharpen For to Screen with an amount of Standard. Use High only for heavily textured products such as leather, knitwear or wood grain.
• Photoshop: after Image Size, apply Unsharp Mask at roughly 60 to 80 percent amount, 0.6 to 1.0 pixel radius, threshold 2 to 4. Judge the result at 100 percent zoom, not fit-to-screen.
• Batch tools: a good resizer applies a mild sharpen automatically after each Lanczos reduction, so you get the benefit without setting it by hand on every file. PixelPrep at pixel-prep.com handles the resize, the sharpening and the per-marketplace dimensions in one pass, which matters when you are pushing hundreds of SKUs a week.

Marketplace recompression: the hit you cannot avoid

Here is the part that catches experienced sellers out. Every major platform re-encodes your image after upload. Shopify, Amazon, Shopee, Lazada and Etsy all run uploads through their own pipeline, which resizes to the sizes their templates need, converts formats on the fly and applies their own JPEG or WebP compression. Your carefully saved quality 90 JPEG becomes whatever quality the platform decides, and you do not get a vote.

Shopify users regularly report that product images uploaded at 2048 pixels or larger are served to shoppers at smaller sizes such as 1800 pixels on the long edge, because the theme requests a specific width from the CDN and the CDN compresses on delivery. Shopify's own guidance is to upload JPEGs optimised to around 85 percent quality and 2048 pixels on the long side, with a hard limit of 20 megapixels and 20 MB per file. Amazon requires at least 1000 pixels on the longest side for zoom and recommends 1600 pixels or more, with zoom quality visibly better at 2000 pixels. Shopee asks for 1024 by 1024 at a 2 MB cap and crops anything non-square.

How to survive the second compression

Because the platform will compress again, your job is to give it the cleanest possible input so the losses stack as little as possible.

1. Upload at the platform's recommended size, not far above it. If you upload 6000 pixels wide, the platform resamples with whatever algorithm it uses, often a fast one, and you lose control of the result. Resizing to the target yourself with Lanczos and a sharpening pass means the platform does only a light touch.
2. Save JPEG at quality 85 to 92. Below 80, the platform's second encode compounds the blocking artefacts. Above 92, the file gets bigger with no visible gain and some platforms simply crush it harder.
3. Use sRGB and strip unnecessary metadata. Embedded Adobe RGB profiles and large XMP blocks get discarded anyway and can shift colours when they are.
4. Avoid PNG for photographs. A 2048-pixel PNG can run to 8 to 12 MB, and the platform converts it to JPEG or WebP on delivery anyway.
5. Keep a master. Store the full-resolution, unsharpened TIFF or high-quality JPEG. Never resize from an already-resized listing image, because each generation adds softness and artefacts.

Retina screens and the 2x rule

Most phones and many laptops now have two or three physical pixels for every CSS pixel. A product image that fills a 500-pixel slot on the page is drawn across 1000 to 1500 physical pixels, so a 500-pixel upload gets enlarged by the browser, and enlargement is the one resize no algorithm can do without softness. The practical rule is to upload at least twice the largest display width your listing uses: for a 600-pixel product image, 1200 pixels is the floor and 2000 pixels is a comfortable target that also enables zoom.

Diagnosing a soft image in 60 seconds

Before reshooting anything, open the master at 100 percent. If it is sharp, the problem is downstream. Then check the upload dimensions, since anything under 1000 pixels on the long side will look soft once the browser scales it. Finally, read the symptoms: a grey, gradual product edge means a soft resampler or no output sharpening, a bright white rim means over-sharpening, rainbow ripples in textures point to a cheap downscaler, and blocky squares in smooth areas are JPEG compression stacking from a low-quality save followed by platform recompression.

Pre-upload checklist

1. Resize from the full-resolution master, never from a previous export.
2. Use Lanczos or bicubic resampling, and avoid any tool that will not tell you which method it uses.
3. Resize to the platform target first: 2048 pixels for Shopify, 1600 to 2000 pixels for Amazon, 1024 pixels for Shopee thumbnails.
4. Apply light output sharpening after the resize, judged at 100 percent zoom.
5. Export JPEG at quality 85 to 92 in sRGB.
6. Confirm the file is under the platform limit: 2 MB for Shopee, 10 MB for Amazon, 20 MB for Shopify.
7. Upload, then view the live listing on a phone and compare it with your export. If the platform's version is noticeably worse, nudge quality up by five points and try again.

Soft product photos are rarely a camera problem. They are a pipeline problem, and a pipeline only needs fixing once. Set the resampler, add the sharpening pass, give the marketplace a clean file, and every image you upload from then on will look as sharp as the one you shot.