How do you get a PDF under a portal's upload limit?
A PDF is shrunk by re-rendering its pages as JPEGs at a lower resolution and rebuilding the document around them. FormworkPDF aims at the size you need — JPEG bytes go roughly with the square of the resolution, so each miss says where to go next — and does it without uploading the file.
Free, no sign-up, nothing uploaded.
Why the limit is always 2 MB and your file is always nine
Upload limits are set by whoever wrote the portal, usually years ago, usually without a scanner in front of them. Meanwhile a scan is a photograph: one A4 page at 300 DPI is roughly 2480 × 3508 pixels, and a scanner saving that with generous quality produces a few megabytes per page before a second page exists. Append a phone photo of an ID and the file doubles again.
So the arithmetic is not on your side, and “compress a bit” is not a strategy. Hitting a hard number is.
How a size target is actually hit
There are two families of PDF compression.
Structural. Remove duplicated fonts, subset what is embedded, drop unused objects, recompress streams. It is lossless, it is polite, and on a scan it saves almost nothing, because a scan is one enormous image per page and the image is already compressed.
Rasterising. Re-render every page as a JPEG at a chosen resolution and rebuild the document around the pictures. This is the only thing that moves the number on a scan, and it is what every tool promising “compress to 2 MB” is doing behind the wording.
The only question left is what resolution to choose, and the answer is aiming rather than guessing. JPEG bytes go roughly with the square of the DPI, so a result that came in at 4 MB against a 2 MB target says fairly precisely where to go next. Long documents get fewer attempts, because the budget is total rendering work rather than a fixed number of tries.
The two things this gets right
It flattens before rasterising. Rendering disables annotations, and an unflattened AcroForm keeps its values in annotation appearance streams. Compress a filled form without flattening it first and the output is exactly the right size with every answer missing — no error, no warning, a perfectly valid empty form. It is the single most expensive silent failure in this whole area.
It refuses to make things worse. Rasterising a compact vector PDF makes it bigger: a page of text is a few kilobytes of instructions and a few hundred kilobytes as a picture. A file already under the target is left alone and says so, rather than being run through the machine and handed back heavier.
What you lose
The text layer. After rasterising, the page is a picture of a page: nothing is selectable, nothing is searchable, and a screen reader gets nothing. That is true of every compressor that hits a hard size target, whatever the marketing says. The dialog states it before it runs rather than behind a disclosure.
If the document has to stay searchable, the alternatives are to send fewer pages, to rescan at a lower DPI in the first place, or to ask the recipient for a larger limit. Sometimes the third one works.
Doing it without the upload
There is a particular irony in a 9 MB scan of your passport being uploaded to a free compression service so that it can be made small enough to upload to the government portal you were trying to reach. The whole operation — render, encode, rebuild, measure, retry — is arithmetic the browser can do, so here it happens in the tab and the file goes nowhere.
Practically, the order that works:
- Open the PDF and fill it in, including any blanks it never declared.
- Append the supporting pages you need — a photo of an ID becomes page seven. Phone photos are rotated by their EXIF tag and capped at about 200 DPI on the way in, which is the cheapest saving available.
- Export. It is flattened.
- Shrink to the portal’s number.
Doing it in that order once beats doing it in the wrong order twice.
Questions people actually ask
- Why is my scanned PDF 9 MB when the paper form was one page?
- Because a scan is a photograph. At 300 DPI a single A4 page is about 8.7 million pixels, and a scanner that saves it with little compression produces several megabytes per page before anything else is in the file.
- Will compressing it make the text blurry?
- At a low enough target, yes — it is a photograph of the page from then on. Aiming at a size rather than a fixed quality setting is what keeps the loss to the minimum that clears the limit.
- Why did my compressed form come back empty?
- Because it was rasterised without being flattened first. Rendering disables annotations, and an unflattened form keeps its values there, so the answers are dropped. Flatten before you compress — here it is forced.
- My PDF is already small and compressing made it bigger. Why?
- A compact vector PDF is mostly instructions, not pictures. Turning those instructions into images almost always costs more bytes than it saves, so a file already under the target is left alone instead.
- Can I still select the text afterwards?
- No. The text layer is destroyed by rasterising — that is true of every compressor that hits a hard size target, and it is said in the open here rather than buried.