Loading...
The upload limit says 5MB. Your file says 41.3MB. Here is what is inside it, why the obvious fixes make things worse, and how to get under the limit without ruining the document.
You have filled in a form, scanned it, and gone to upload it. The site wants 5MB. Your file is 41.3MB. You try "Save As", which changes nothing, and then you email it to yourself in three parts like it is 2004.
The frustrating part is that the document is four pages of black text on white paper. So where did forty megabytes come from?
PDF is a container. What is inside it makes an enormous difference, and there are two very different kinds of file wearing the same extension.
A PDF exported from a word processor contains text — actual characters, with a font, positioned on a page. Four pages of that is perhaps 80KB. You can select the words, search them, and copy them out.
A PDF scanned from paper contains photographs. Each page is a picture of a page, and your scanner very likely took it at 600 dots per inch in full colour, because that is what scanners default to. An A4 page at 600 DPI is roughly 4,960 × 7,015 pixels — about 35 megapixels, per page. That is a larger image than most cameras produce, and you have four of them.
Forty megabytes is not a bug. It is exactly what you asked for.
Open the file and try to select a line of text with your cursor.
This distinction decides everything that follows, which is why nearly every "compress PDF" tool gets the job half right: it treats both the same.
Re-saving the file. Opening a PDF and saving it again mostly rewrites the same data. Some viewers remove unused objects and save you two percent. That is not the problem.
Zipping it. PDFs already compress their contents internally, and JPEG images inside them are compressed too. Zip finds almost nothing left to squeeze. A 41MB PDF zips to about 40.6MB, and now the recipient has to unzip it.
Printing to PDF. This sometimes works, by accident, because the printer driver rasterises everything at screen resolution. It also destroys your text layer without telling you. More on why that matters below.
Re-render the pages at a sensible resolution. 600 DPI is archival quality — appropriate for photographing a painting, absurd for a signed form. 150 DPI is sharp on any screen and prints perfectly well. Dropping from 600 to 150 removes three quarters of the pixels in each direction, which is about a sixteenth of the data.
That single change usually takes a 40MB scan to under 3MB. You can do it with the PDF compressor, which does exactly this in your browser — the file never leaves your device, which matters when the thing you are compressing is a bank form.
Colour is the other lever. A scan of black text on white paper stored in full colour is carrying three channels where one would do.
Do not rasterise it. Compressing a text PDF by turning its pages into images is a bad trade twice over: it usually makes the file bigger, because a picture of a page is larger than the instructions to draw it, and it destroys the text layer.
Instead, find the images inside it. A twelve-page report that weighs 30MB almost always contains three photographs somebody dropped in straight from a phone. Replace them with resized versions at the size they actually appear on the page and re-export.
The fastest way to make a PDF smaller is to send less of it. If the form is on pages 3 and 4 of a 40-page pack, extract those two pages. It takes ten seconds, the result keeps its text layer intact, and it is more considerate to whoever has to read it.
When a page becomes an image, the text stops being text. It cannot be selected, copied, or found with Ctrl-F. And crucially, it cannot be read by software.
That last point has consequences beyond convenience. Employers run applications through systems that read the text out of your file. A résumé exported as a picture of itself is a blank page to that software, and you will never be told that is why you heard nothing back. The same applies to any document that will be processed rather than read: invoices going into an accounts system, forms going into a case management tool.
So: rasterise scans, which were images already and lose nothing. Never rasterise a document whose text matters.
| What you have | Do this | Expect |
|---|---|---|
| Scan at 600 DPI | Re-render at 150 DPI | 90%+ smaller |
| Scan in colour, text only | Re-render at 150 DPI | Smaller again |
| Text PDF with photos | Resize the photos, re-export | Large reduction |
| Text PDF, already lean | Send fewer pages | Proportional |
| Anything | Zip it | Almost nothing |
Set your scanner to 200 DPI and greyscale as the default, and change it upward on the rare occasion you are scanning a photograph. Every document you scan from then on will be a tenth of the size and completely legible, and you will never think about this again.