Visa portals, job applications, university forms, driving licence renewals, tax filings. Somewhere in each of them is a number — 200 KB, sometimes 100, occasionally a genuinely cruel 20 — and a photo four or five times too big for it.
The usual advice is to drag a quality slider until the number goes down. That works, eventually, and it is the reason so many uploaded photos look like they were printed on a wet newspaper. There is a better order.
File size is not a setting
You cannot type 200 KB and have a picture obey. The size that comes out is the result of three separate things:
- How many pixels there are. A 12-megapixel phone photo has twelve million of them to describe.
- How hard they are compressed. The quality setting — how much detail the encoder is allowed to throw away.
- How complicated the picture is. The one nobody mentions, and often the biggest surprise.
That third one is why the same settings give wildly different results. A photo of a plain painted wall compresses to almost nothing: there is barely any detail to store. A photo of a tree in leaf, at identical dimensions and identical quality, can be five times larger, because every leaf is detail the encoder has to describe.
So “save at quality 80” is not a size, and never was. It is an instruction whose result depends on the photograph.
The order that works
1. Cut the pixels first. This is the lever with the best exchange rate, and almost everybody skips it.
Halving the width and the height leaves a quarter of the pixels. A 4000-pixel-wide photo shrunk to 1000 wide has one sixteenth as many, and the file falls roughly in step. If the photo is going to be looked at on a screen inside a web form, 1000 to 1600 pixels on the long edge is already more than will ever be displayed.
Compare that with the alternative: to get the same saving from quality alone, you would have to compress so hard the picture visibly falls apart. Resizing costs you detail nobody was going to see. Over-compressing costs you detail everybody can.
2. Crop what is not needed. Fewer pixels again, and this time you choose which ones go. A document photographed on a desk is usually half desk. An ID photo does not need the room behind you.
3. Then, and only then, lower the quality. With sensible dimensions in place, a moderate quality setting will usually land under the limit on its own, and the damage is spread over a picture that no longer has detail to spare.
4. Check the result at full size before you send it. Not the thumbnail. Open it, look at the smallest thing that matters — a date, a serial number, the edge of a signature — and decide whether it survived.
How a “compress to 200 KB” tool actually works
Worth knowing, because it explains what it can and cannot do for you.
There is no formula that turns a target size into a quality setting. The encoder’s output depends on the picture, so the only way to hit a number is to try: encode at some quality, measure, and adjust. A tool doing this properly runs a binary search — too big, go lower; fits, try a little higher — until it finds the highest quality that still fits under your number.
And when quality alone cannot get there, which happens constantly with small targets, it has to reduce the dimensions as well. A twelve-megapixel photo cannot be squeezed into 20 KB by any quality setting that leaves a recognisable picture. Something has to give, and pixels are the thing that gives most gracefully.
That is exactly what the size pages here do — the target is filled in, the search runs on your own machine, and the result is shown before anything is saved.
When the limit cannot be met
Sometimes the number is simply too small for the photo, and there is no clever setting. Three things to try, in order:
- Crop harder. A tighter frame is fewer pixels and less complexity at once, and for an ID photo it is usually what the form wanted anyway.
- Check what format the form accepts. If it takes PNG and you are sending a photograph, you are fighting with one hand tied — JPG is several times smaller for a photograph. If it takes WebP, smaller still.
- Ask. Limits on old forms are frequently inherited from a system nobody has looked at in fifteen years, and the office behind it often accepts an emailed file without blinking.
The trap: identity documents
Worth saying plainly, because it costs people weeks.
Passport pages, ID cards, certificates and bank statements are read by somebody — or something — looking closely at small printed detail. A compressed scan that looks perfectly fine at screen size can be rejected as illegible when a reviewer zooms in on a document number, and the application goes back to the end of the queue.
For anything being used as proof of identity, resize and crop, but keep the quality high. If the file still will not fit, that is a conversation with the office, not a reason to compress until the numbers blur. The same logic applies to a PDF of the same documents.
Two things that make it worse
Compressing the same file over and over. Each pass compresses the previous pass’s damage along with the picture. If the first attempt missed the target, go back to the original and try again from there — never compress the compressed copy.
Turning a screenshot into a JPG to make it smaller. It usually does not work, and it always looks worse. Screenshots are hard edges and flat colour, which is what JPEG damages first and what PNG compresses best. Resize the screenshot instead, or crop it to the part that matters.
Why these forms ask for so little
Not to torture you, though it is a reasonable suspicion.
Most of these limits are old. They were set when storage was expensive and a government department was budgeting for tens of thousands of applicants, on connections that made a 4 MB upload a genuine problem. The number stayed after the reason for it went away, because the form works and nobody reviews a working form.
Which is also why arguing with the limit rarely helps, and why a photo that is 1200 pixels wide instead of 4000 is not a compromise — it is simply the right size for the job.