Social Media Image Size Guide — published sizes and what a crop costs
Two different questions hide inside “what size should this image be?”. The first is a matter of record: what dimensions does a platform actually publish for a given placement? The second is geometry: given the image you already have, what happens to it when it meets that frame? This tool keeps them apart. The reference table answers the first and cites a source for every figure. The fit and crop calculator answers the second, and needs no table at all.
A display size is not a hard limit
These two constraints get flattened into one number constantly, and the difference decides whether your upload fails or quietly changes. A hard limit is a rule the upload is rejected for breaking: a minimum width, a maximum file size in bytes, a format that is not accepted. You find out immediately. A display size is the frame the image is shown in after it has been accepted. Hand over something a different shape and the platform scales and crops it to fit on its own, without telling you which pixels it dropped — which is how a portrait photo ends up as a wide banner with the top of someone’s head missing. Each row in the table records which kind of constraint it is, taken from the same page as the numbers.
Aspect ratio matters more than absolute pixels
Resolution is negotiable; shape is not. An image at 1280×720 and one at 1920×1080 are the same shape — both reduce to 16:9 — so moving between them only resamples the picture. Every pixel that was in the frame is still in the frame. Change the ratio, though, and something has to be thrown away or something has to be added, because there is no way to map a tall rectangle onto a wide one while keeping all of it and filling all of it. That is why the calculator reduces both sizes to their ratio first, with a greatest-common-divisor reduction, and falls back to a decimal form like 1.91:1 when the whole-number terms get too large to recognise.
Fit and fill are the two honest answers
When the shapes disagree, you have exactly two options and both cost something. Under fit, the image is scaled until the whole of it sits inside the target. Nothing is lost, but the target is not filled: the leftover shows as empty bars above and below (letterboxing) or at the sides (pillarboxing). Under fill, the image is scaled until it covers the target completely and the overhang is cut off. The frame is full, but pixels are gone — and because the crop is centred, what goes is whatever sat nearest the edges. The calculator reports the scaled size, the depth of each bar or each cut in pixels, and the share of the original area a fill discards, so “it gets cropped a bit” becomes a number you can decide against.
Why the diagram is the point
A percentage is hard to picture. The diagram draws both rectangles on a single shared scale around one centre, so the overhang is visible as an overhang: you can see the band that a crop removes, and its depth is written on it in pixels. Every shaded region is named in the legend in words as well, so nothing in it depends on being able to distinguish the shading.
Enlarging does not add detail
Why every row carries a link and a date
A dimension is a fact about someone else’s product, and platforms change published sizes, aspect ratios and file-size caps without notice and without a changelog. Aggregator pages repeat figures that were already wrong when they were copied. So a row is only allowed to exist here with the address of the page the figure was read on and the date it was read, and the table shows you both. Treat each one as a snapshot on its own date and nothing more: before you export something that matters, open the link and confirm the number still reads the same. None of this data is live, fetched or synced, and it is deliberately better for the table to be short than for it to be confident and wrong.
What this tool does not do
It never touches your image. There is no canvas drawing, no re-encoding and no modified file to download — it is arithmetic and an SVG diagram. If you point it at a local file, the only thing read is the width and height, purely so you do not have to type them; the file never leaves your device. Rotated JPEGs are decoded through the browser’s orientation-aware path so the numbers match what you see rather than how the bytes happen to be stored, and the tool says so when that path is not available.