Etsy & Amazon Image Resizer

Fit or Fill: choosing per marketplace

The two modes differ in one line of arithmetic. Fit takes the smaller of the two scale factors needed to reach your target, so the whole image lands inside the frame and the leftover space is filled — with white, with a color you pick, or with nothing at all if you choose Transparent and export PNG. Fill takes the larger factor, so the image covers the frame completely and whatever hangs over the edges is discarded. Fit always centers the artwork and gives you no say in it; Fill is the mode with the 3×3 grid, because once you are throwing pixels away it matters which ones.

Take a portrait product shot at 1600 × 2400 px and send it to the Amazon Main image preset, 2000 × 2000 px. Fit compares 2000 ÷ 1600 = 1.25 against 2000 ÷ 2400 = 0.833 and takes 0.833, so your photo is drawn at 1333 × 2000 px — full height, and (2000 − 1333) ÷ 2 = 333 px of padding down each side, none top or bottom. Fill takes 1.25 instead and draws the photo at 2000 × 3000 px, which is 1000 px taller than the frame, so with the center anchor 500 px are cut off the top and 500 off the bottom. Measured back on the original that is 400 rows from each end — 800 of your 2400, a third of the photograph gone.

For an Amazon main image that arithmetic settles it: Fit with White background. Amazon wants the product on pure white with nothing else in frame, and the padding the tool adds is the same pure white, so the two halves of the canvas are indistinguishable — you get a compliant square without cropping the product's feet off. An Etsy lifestyle shot wants the opposite. Nobody scrolling a search grid rewards you for showing the whole photo; they reward a frame that is full edge to edge, and the 3000 × 2250 Listing photo preset with Fill gives you that. Move the anchor off center if the subject sits high — a top-center anchor keeps a hanging plant or a head, a bottom-center anchor keeps a table surface.

The case worth naming is the one where neither answer is good — and the example above is exactly it. Fit leaves the product marooned in a frame it does not fill; Fill throws away a third of the shot. No setting rescues a source whose aspect ratio is that far from the target. Re-crop by hand so the subject is composed for a square, or re-shoot with the square in mind. A resizer can only decide which pixels to keep; it cannot decide what the picture should have been.

What a resize costs you

Downscaling is nearly free. Going from 4000 px to 2000 px, the browser has four source pixels to average into every output pixel, and with smoothing set to high the result is often crisper than the original looked at that size. Upscaling is a different act entirely. Pushing an 800 px photo to the 2000 px Amazon preset is a 2.5× enlargement in each direction, which is 6.25 output pixels for every source pixel — more than five in six of them invented — and no interpolation filter in any browser recovers detail that was never captured; it can only make a confident guess about the gaps. The print-readiness panel flags this as you work. It reads the scale the image is actually drawn at rather than the size of the target box, so a Fit into a frame wider than your source still reads as no upscaling when the fit works out to a shrink, and the badge turns from amber to red once that drawn scale reaches 2×. The cruelty of it is that the softness lands precisely where the marketplace's zoom viewer magnifies it: the shopper who cares enough to hover is the one you show the mush to.

Re-encoding has a quieter cost. The format dropdown opens on JPEG with quality at 90, so unless you change it your download is a lossy encode. That is fine once. Do it to a file that was already a JPEG and it is the second time those pixels have been through the quantizer, and the artifacts of the first pass — ringing along hard edges, blocking in smooth gradients — are now real image content that the second pass faithfully preserves and adds to. Nothing looks broken after one round trip, which is why people keep going.

So the discipline is simple: keep the camera original, and produce every marketplace size from that one file rather than resizing a resize. The batch panel already works this way — each preset is drawn from the same decoded source into its own canvas, so five outputs are five independent single-generation encodes, not a chain. And when a file is going to be edited again later, switch to PNG; the quality slider disappears because there is nothing left to trade away. It matters most when the same artwork also has to print: a listing photo gets away with 2000 px, but a print file is sized from the product's print area, which is what the print-readiness readout is really telling you and what our guide to print-on-demand image requirements works out per product.