Background isn't pure white
Not "light grey". RGB 255, 255, 255. Supplier photos shot on a countertop or a softbox gradient fail this, and hand-painted white leaves halos around the edges.
A background that isn't pure RGB 255 white. A product that fills less than 85% of the frame. A longest side under 1600 px. Or text, logos and props burned into the shot. ImSpec fixes the first three across a whole restock in one pass — white replace, 92% fill on a 1:1 canvas at 2000 px, weight under the cap — then audits every file before you upload. It runs in your browser; the photos never leave your laptop.
Three can be measured on a pixel grid, so a machine can fix them and re-check them. The fourth needs eyes — this tool will not pretend to see text.
Not "light grey". RGB 255, 255, 255. Supplier photos shot on a countertop or a softbox gradient fail this, and hand-painted white leaves halos around the edges.
A 2000 px canvas with the product floating in the middle reads as a small image; Amazon's guidance is 85% or more. Most supplier shots sit at 30–50%.
Main images go on a 1:1 canvas with the longest side at 1600 px or more, commonly exported at 2000. A 4:3 photo padded badly leaves white bands; a 12 MB PNG gets refused.
Main images are supposed to be the product on white — no promotion copy, no badges, no extra objects. This is the one a pixel check cannot judge.
Requirements change; check Amazon's current image guidance before a launch. What is fixed here is the mechanical part — colour, fill, shape, weight, naming — not the editorial judgement.
No account, no project setup. The page opens straight into the queue, and the settings you used last time are still there.
Drag a folder's worth of JPG, PNG or WEBP in, or pick from disk. Every file is audited on import, so a 4:3 photo or a 9 MB PNG surfaces before you process anything.
Amazon main image is pre-selected. Whiten, re-centre at 92%, square to 2000, cap the weight — then compress, rename and check in the same queue. Outputs stack on the same image instead of scattering into folders.
One ZIP, files numbered IMG-001, IMG-002 in the order you dropped them, plus an audit CSV for the SKU sheet. Upload and go.
Main images are the reason to buy. These are the reasons to keep the page open every restock.
Type 120 KB and the encoder finds the quality and long side that fits, then tells you where it landed. For gallery shots and slow-loading listing pages.
One prefix, a start number, digit count, separator — plus a two-column CSV so filenames and listing rows stop drifting apart.
Text, corner or centre, size and opacity. Not for main images — text is exactly what gets a main image suppressed.
2 × 2, 3 × 3 or 1 × 2 with cell size, gap and background. For variation listings and ad creatives.
Five rule sets including Amazon main and secondary. Per image: squareness, border whiteness, subject fill, weight, alpha channel, format — pass, warn or fail, with the reason behind each flag.
No AI cut-out for hair, glass or reflections. No text detection. No beautifying. Those need a real masking tool or your eyes — pretending otherwise is how a seller ends up with a rejected batch.
Every value below is editable in the app — these are just the defaults the Amazon preset ships with.
| Property | Value | Why |
|---|---|---|
| Canvas | 2000 × 2000 | 1:1, comfortably above the 1600 px longest-side floor |
| Fill | 92% | Clears the 85% guidance without touching the edges |
| Background | RGB 255 · 255 · 255 | Pure white, filled before the product is drawn |
| Weight | under 10 MB | Main-image ceiling; JPEG quality walks down to fit |
| Format | JPEG | No alpha channel, photo-safe, universally accepted |
| Naming | IMG-001, IMG-002… | Order matches the order you dropped the files in |
The audit reads pixels, not intent: it cannot tell whether a logo is printed on the product or stuck on as a badge. Use it to catch mechanical mistakes, then look at the image yourself.
No subscription, no per-seat pricing, no credits. The free tier is the real tool with a batch cap — not a video demo.
Yes. The canvas is filled with #fff before anything is drawn, the near-background pixels are rewritten to 255, and the alpha channel is dropped on export. What you get back is an opaque white JPEG, not a transparent PNG that shows grey in a viewer.
Pixels close to the sampled border colour are pushed to white, so a cast shadow on a white backdrop usually disappears. For a main image that is what you want; if the shadow is part of the look, don't run that shot through the whiten step.
You can, until the next supplier ships photos with a different background colour and different framing. Actions replay one recipe; these tools measure each image — border colour, bounding box, weight — and warn you when a photo is too messy for an automatic pass.
That is the Pro tier: 200 per batch, every marketplace preset in one run, one ZIP at the end. The free cap exists because the batch is the product.
No. Every tool runs in the browser tab using canvas — no server, no account, no tracking pixel. Once the page has loaded you can disconnect and keep working.
Ask for a refund within 14 days. One email, no form.
Open the toolbox, run three of your own main images through it, and decide with the audit CSV in front of you.