Etsy 20MB File Fitter

What actually happens when the file is too big

Nothing dramatic, which is the problem. Etsy caps each file attached to a digital listing at 20 MB, and a listing can carry five files at most. Go over on one of them and the file is simply refused — there is no warning that you were close, no offer to compress it for you, and no partial upload. You are left staring at an export that took forty minutes and a listing you cannot publish. Those are the limits as they stand in July 2026; check your seller dashboard before you build a product around them.

The five-file allowance is the part people forget, and it is usually the easiest way out. A 40 MB bundle that will never fit as one file fits comfortably as three. There are two catches worth knowing before you split. First, the name itself: it must be 3 to 70 characters and may contain only letters, numbers, hyphens, underscores and periods — a space or an ampersand is rejected at upload, so Wall Art & Print.png never gets in the door. Second, the name you do settle on is shown to the buyer exactly as you typed it and cannot be renamed afterwards; Etsy has no edit-name control, only delete and re-upload. So final_v3_USE_THIS_ONE.png is a decision you make once and live with. Digital listings also do not support variations, so "pick your size" has to be five files or five listings, not a dropdown.

The trap is reaching for a quality slider before you know which problem you have. "Too big" is a symptom with at least four different causes, and three of them are not fixed by compression at all.

Why quality 85 usually looks the same as quality 100

JPEG does not store pixels. It chops the image into 8 × 8 blocks, converts each one into a set of coefficients describing patterns at different scales — broad gradients through to detail that changes every other pixel — and then divides those coefficients by a table of numbers before storing them. That division is the whole trick. Coarse division throws away precision; fine division keeps it. The quality setting is simply how coarse that table is.

At the very top of the scale the table is close to all ones, which means the file is spending bits preserving distinctions your eye cannot resolve: the difference between two nearly identical shades in a patch of sky, at a scale of two pixels. Step down to around 85 and the coarsening lands almost entirely on the highest-frequency coefficients first — the ones carrying that invisible detail. Most encoders also start averaging colour information across neighbouring pixels around there, which human vision tolerates far better than it tolerates blurred brightness. The result is a file roughly a third the size that is genuinely difficult to tell apart from the original at any normal viewing distance.

That is why the search on this page never goes above 0.95. The last few points of the quality dial are the worst value in the whole range — they can double a file without changing anything a buyer will see. The search also refuses to go below 0.4, because past that the block structure starts showing as visible squares and halos, and at that point scaling the image down is a better trade than crushing it further. Everything between those two bounds is found by binary search, and every candidate is a real encode, so the size reported is measured rather than estimated.

Why your PNG is enormous, and when it is still the right choice

PNG is lossless: it guarantees that the pixels you get back are bit-for-bit the pixels you put in. It earns its size back by predicting each pixel from its neighbours and storing only the difference. On flat-colour artwork the prediction is nearly always right, the differences are almost all zero, and long runs of zeros compress to almost nothing. That is why a poster of solid shapes can be a couple of megabytes at 6,000 px.

Photographs break the prediction completely. Every pixel of a photograph carries a little sensor noise, so the difference between a pixel and its neighbour is essentially random, and random data does not compress. A photograph saved as PNG is close to storing raw pixels, which is why a 6,000 × 4,000 px image lands somewhere around 60 MB while the same image as a quality-90 JPEG is under 5 MB and looks the same.

The reverse case is just as sharp. Hard edges — black text on white, crisp vector-style line art, a two-colour logo — are exactly what JPEG's block transform handles worst, producing faint grey halos and ringing along every boundary. On a printable that is mostly typography those artefacts are the first thing anyone notices. That is the case where PNG is correct and you should shrink the pixel dimensions instead. If your file needs transparency, WebP is the compromise: an alpha channel and a real quality dial in the same format. Bear in mind that JPEG has no alpha channel at all — when you select JPEG here, this tool lays down a white background before drawing, because transparent pixels would otherwise come out black.

Re-saving a JPEG is not free

Every time a JPEG is opened and saved again, it is decoded back to pixels and then re-quantised. The second pass does not know which values the first pass rounded, so it rounds again on its own grid, and the two sets of errors do not cancel — they accumulate. Do it half a dozen times and the halos around edges thicken visibly, even if you chose "maximum quality" every single time. This is generation loss, and it is the reason a file that has been through three rounds of edits looks worse than a fresh export at a much lower quality setting.

It matters here because this tool always decodes and re-encodes: feeding it a JPEG costs you one generation, unavoidably. The habit that avoids the problem entirely is to keep your master in a lossless format — PNG, TIFF, or your editor's native file — and export a fresh JPEG from that master whenever you need one. Never edit the JPEG you exported yesterday. If you are preparing files for a print provider rather than a buyer, their input requirements usually differ from Etsy's, and our notes on print-on-demand image requirements cover what they normally ask for.

File size and resolution are different problems

These get confused constantly, because both get called "size". A file's size in megabytes is how many bytes it takes to store. Its resolution is how many pixels wide and tall it is. They are related — more pixels usually means more bytes — but the relationship is loose enough that you can halve one without touching the other. Compression changes bytes and leaves pixels alone. Scaling changes pixels, and bytes follow roughly with the pixel count, which is why halving the long edge typically cuts the file to a quarter rather than a half.

Then there is DPI, which is neither. The DPI number stored in a file is a note to the printer about how large to print the pixels; changing it from 72 to 300 without resampling adds not one byte and not one pixel. What determines whether your buyer gets a sharp A3 print is the pixel count, full stop — what DPI and PPI actually mean untangles this properly. The practical consequence for the 20 MB limit is that shrinking the long edge is the honest lever: it reduces the file and the printable size together, so you know exactly what you are giving up, whereas compression reduces the file by quietly degrading detail you may not notice until someone prints it at A2.

What to do when your file is still too big

• Prefer JPEG for flat artwork and photos. It is the only format here with a real compression dial, so it can land almost exactly on your target. PNG is lossless — the only way to make a PNG smaller is to shrink its pixel dimensions, which is why this tool falls back to scaling when you pick PNG.
• Ship a ZIP instead of a giant image. The per-file cap applies to each file you attach, so a bundle that is too heavy as one enormous PNG often fits comfortably once it is split into a few files, or zipped and attached as a single archive.
• Give buyers the print size they will actually use. A 300 DPI file sized for an A2 print is enormous; most buyers print at A4 or smaller. Offering a couple of sensible sizes beats one unmanageable master file.
• Watch the long edge before the quality slider. Halving the long edge cuts the pixel count to a quarter, which usually saves far more than dropping quality from 90 to 70 — and it looks better at normal viewing sizes.

How this fitter picks a quality level

Most compressors give you a slider and leave you to guess. That is backwards: you do not care what the quality number is, you care that the file fits and still looks good. So this tool works from the target backwards. It re-encodes the image at the highest quality it is allowed to use, and if that is already under your target it stops there and hands the file straight back — no quality thrown away for nothing.

When the first attempt is too heavy, the tool runs a binary search across the quality range. Each round it encodes at the midpoint, keeps the result if it fits, and narrows the range toward whichever side is still unexplored. Eight rounds is enough to pin the quality down to roughly one part in 250, which is far finer than the eye can tell apart. Every candidate is a real encode, so the size you see is measured, not predicted from a formula.

Some files cannot be squeezed under the target by quality alone — a very large PNG with transparency, or a target of a couple of megabytes on a 40-megapixel photo. In that case the tool scales the image down in steps and retries, so it always converges on something that fits rather than reporting failure. The results table shows you exactly what it did: the final file size, the dimensions it ended up at, and the quality setting it settled on, next to the original for comparison.

Frequently asked questions