How to get a PDF under 100 KB for an online form

The form says maximum 100 KB. Your scanned marksheet is 2.4 MB. You compress it, upload it, and the portal rejects it again — or worse, accepts it, and three weeks later the verification officer cannot read your roll number.

This is one of the most common ways an application goes wrong, and almost nobody explains what is actually happening. A scanned page is not a document to your scanner. It is a photograph of a document, and photographs are large for reasons that have nothing to do with how much text is on the page.

Why a single page comes out at 2 MB

Three settings decide the size of a scan, and page count is not one of them. Resolution, colour depth and format between them account for almost the entire file.

An A4 page scanned at 300 DPI is roughly 2,480 by 3,508 pixels — about 8.7 million of them. In full colour, each pixel carries three bytes, so the raw image is around 26 MB before any compression. JPEG brings that down to somewhere between 1 MB and 3 MB depending on quality. That is your 2.4 MB file, and it is the scanner default that produced it, not anything you did.

The pattern is worth reading carefully. Dropping from colour to greyscale removes roughly two thirds of the file and costs you nothing on a document that was printed in black ink. Halving the resolution removes three quarters again, because pixel count scales with the square.

The mistake that gets applications rejected

Faced with a 100 KB limit, most people take the 2.4 MB colour scan and push a compression slider until the number goes green. That works arithmetically and fails in every other way.

JPEG compression does not remove resolution. It removes detail within the resolution it was given, and it removes it from the highest-frequency parts of the image first. In a scanned document, the highest-frequency parts are the edges of the letters. Crush a 300 DPI colour scan to 100 KB and you keep every one of those 8.7 million pixels, each now a smeared approximation of what was there. The result looks acceptable at thumbnail size and turns to mush the moment anyone zooms in to read a date.

A file that meets the size limit and cannot be read at full zoom has not solved the problem. It has moved it to someone who cannot contact you about it.

The correct order is the opposite. Reduce the resolution and the colour depth first, so there are fewer pixels to describe, then compress gently. A 200 DPI greyscale scan compressed lightly to 100 KB carries far more readable detail than a 300 DPI colour scan crushed to the same number, because the compressor was not asked to throw away as much.

What to do, in order

1. 1Rescan if you can. Set the scanner to greyscale and 200 DPI before it ever produces a file. This single step usually lands you between 250 and 500 KB, and the quality is genuinely better than compressing a colour scan to the same size.
2. 2Crop the margins. Scanners capture the full platen, including the white border and often a strip of the lid. On a certificate that is easily a fifth of the pixels, describing nothing.
3. 3Convert to greyscale if the original was colour but the content is not. A marksheet with a coloured seal still reads perfectly in greyscale, and most portals only require the seal to be visible, not chromatic.
4. 4Compress last, and only as far as you need. Aim to land just under the limit rather than as far below it as possible — there is no prize for 40 KB.
5. 5Open the result at 100% zoom and read the smallest text on it. If you cannot, neither can the person verifying it.

When the document is a photograph, not a scan

Most people now photograph documents with a phone rather than scanning them, and phone cameras make the problem worse in a specific way. A 12-megapixel photograph of an A4 page is around 4,000 by 3,000 pixels, which sounds similar to a 300 DPI scan. The difference is that a scan is evenly lit and perfectly flat, and a photograph is neither.

Uneven lighting creates gradients across the page. Gradients are expensive to compress, because the compressor cannot treat the background as one flat colour. A photograph of a white page is often larger than a scan of the same page at the same resolution, purely because the white is forty slightly different whites.

If you are photographing, use your phone's document mode where it exists. It flattens the perspective, evens the lighting and usually converts to greyscale, which is most of the work done before the file is created. Shoot in daylight, not under a ceiling light directly overhead, which throws a shadow of your own hands onto the page.

Common limits and what they mean in practice

If a portal demands 50 KB for a multi-page certificate, it is asking for something that cannot be delivered at readable quality, and the practical answer is to submit the cleanest black-and-white version you can produce and keep the original for the verification stage. Do not spend an hour trying to fit a five-page document into a limit designed for one.

What none of this fixes

Compression cannot recover detail that the scan never captured. If the original photocopy is faint, or the certificate has been folded through a line of text, no amount of processing will make that legible — it will simply make a smaller file containing the same illegible mark. Get a better source document before you spend time on the file.

It also will not fix a rejection caused by something other than size. Portals commonly reject on dimensions, on format, on colour mode, and on file name characters, and almost all of them report every one of those as a generic upload error. Before assuming the size is wrong, read the instruction page again for a required width in pixels or a specific format, because those are the constraints people most often miss.

And keep the original. Every compressed copy is a lossy copy, and compressing a compressed file again compounds the damage rather than restarting from the source. Store the full-quality scan somewhere you can find it, and produce a fresh reduced version each time a portal asks for a different limit.