← All posts

Why won’t my PDF get any smaller when I compress it?

You put a 9 MB PDF into a compressor. It thought about it, and handed you back 9 MB. Or 8.7 MB, which is worse, because now you have to decide whether it is worth trying again on the hard setting.

So you tried the hard setting. Same number. Then a different website, then another one, and by now you suspect either the file is cursed or every compressor on the internet is a fake.

Neither. The file is telling you something specific about what is inside it, and once you know what that is you will know within a minute whether there is anything left to win.

Why a PDF stops getting smaller

A PDF is not one lump of data that can be squeezed. It is a bundle of separate things, and only some of them respond to compression at all.

The text is already compressed, and it was never big. A page of text is not a picture of words — it is a short list of instructions saying which glyph goes at which coordinate, and that list is already Flate-compressed inside the file. A dense page of type comes to a few kilobytes. Two hundred of them come to a few hundred kilobytes. There is no version of squeezing that turns that into something meaningfully smaller, because it is already about as small as text gets.

The fonts are stored once each. An embedded font is 100–400 KB, stored once for the whole document rather than once per page. It can be a large share of a small file, and there is nothing to re-encode there either.

Which leaves the images. This is what compression actually operates on. A compressor looks at each picture, works out how big it is drawn on the page, converts that into a resolution, and re-saves anything above the ceiling at a lower one. That is where the megabytes come from and that is where they go.

So there are exactly two reasons your file did not move:

  1. There are barely any images in it. A text report with two logos has nothing to give up. A 400-page contract is large because it is 400 pages, not because anything in it is heavy, and no setting on any compressor changes that.
  2. The images are already efficiently encoded. A photograph that is already a JPEG at 150 dpi is already at the ceiling. Re-saving it at a similar quality produces a similar number of bytes, and occasionally a few more, because a second lossy pass is never a free copy — why a re-encode can come out heavier is a post of its own.

Why getting your original file back is the correct answer

If re-encoding would produce a bigger file than the one you uploaded, Compress PDF hands you your original bytes back and says so in the notes. It does not quietly give you the larger result and call it compression.

That looks like failure and it is the opposite. A tool that insists on returning something hands you the loss instead, and you do not notice until you check the byte count. Same size means "already as small as this can make it", not "the button did not work".

The notes are the useful part. They say how many pictures were re-saved, how many were already at or below the target resolution, and how many could not be touched. If it says there are no pictures in the document at all, you have your answer and can stop trying.

Find out what is actually taking up the space

Before you try a fourth setting, spend two minutes working out what you are dealing with.

  1. Check the basics. Our PDF inspector runs in the browser without uploading anything and reads out the page count, the page sizes, and — the important one here — whether the document has a real text layer.
  2. Divide the size by the page count. Under about 50 KB a page, it is a text document and it is already small; the size is coming from the number of pages. Over about 500 KB a page, it is a scan or it is full of photographs, and there is plenty to win.
  3. Pull the pictures out and look at them. Extract images from PDF gives you every image in the file at the pixel size the document actually stores, in one zip. Sort by file size. If the top three files account for most of the document, you have found your problem. If the zip is nearly empty, there is nothing to compress and you have your answer.
  4. Run Compress PDF once at the default — balanced, 150 dpi — and read the notes rather than only the number. "12 pictures were already at or below the target resolution" means a harder setting will not help much either.

The one thing that shrinks almost any scan a lot

If the file is a scan and it is in colour, tick turn pictures grey in the compress options before you reach for the hard quality setting.

Office scanners default to colour on paper that is black text on white. That stores three channels where one would do, and it also defeats JPEG, because the sensor noise across a "white" page is coloured speckle no encoder can find a pattern in. On a colour scan of a black-and-white original, dropping to one channel routinely saves more than the quality slider can, and shows less.

It is one-way. A grey scan does not remember what colour it was, so keep the original if a red stamp or a highlighted clause might matter later.

When your "text" PDF is secretly a picture of text

This is the case that confuses people most. The document looks like a letter, it is 40 MB, and nothing compresses it — because it is not a letter, it is photographs of a letter. Every page is one big image with no text behind it. Try to select a word: if the cursor sweeps across the page without highlighting anything, there is no text layer.

Our OCR PDF tool tells you which pages are pictures and which carry real text. Read this before you upload, though: it does not add a text layer. Tesseract is not installed here, so the file you get back is the file you sent, and a report is all you gain — which pages actually need recognition goes through what that report can settle. Useless if you wanted searchable output; for that you need something local, see below.

The good news is that a scan is all images, so the dpi ceiling and the greyscale option are the right levers and they will work. A 300 dpi colour scan typically loses half to three quarters of its size.

What this will not do

It never rasterises the pages. Turning a document into pictures of itself will shrink almost anything, and it destroys the text — no selecting, no searching, no copy and paste, nothing for a screen reader. Other tools offer that trade silently. This one does not offer it at all. If you genuinely want it, and the document is disposable, PDF to JPG renders each page as an image at a resolution you choose and JPG to PDF puts them back into a PDF. Know what you are giving up before you do it.

It does not subset fonts, deduplicate repeated images or remove pages. If your file is mostly fonts, or mostly pages, compression is not the tool.

It cannot open an encrypted file. Unlock it first, with the password you already have.

The detail thrown out of a re-saved picture is gone. Keep the original if the photographs matter. Related: if the images you pulled out in step 3 looked soft, that is usually not the extraction's fault — they were stored that way.

When the size does not add up at all

Occasionally a file is large for a reason unrelated to its contents. A form saved thirty times keeps a complete copy of every one of those saves, and none of them shows up as a page or a picture, so nothing above will find it. It is also what makes such a file grow when you try to compress it.

Repair PDF is the tool for that. A rebuild keeps only what the document still points at, so on a much-edited form it can take out more than any compressor, and it does nothing at all to a healthy file. It costs you any digital signature — what a repair changes, and by how much.

When a desktop tool is genuinely better

If the browser tools have told you the file is already as small as it gets and you still need it smaller, these are the honest next steps, and all three are free:

If it truly will not get smaller

Then stop trying to compress it and change the problem. If the ceiling is an upload limit, Split PDF by size cuts the document into parts that each stay under a size you set, which is what most portals actually want. If the ceiling is an email attachment, send a link instead.

A 400-page text document at 12 MB is not a badly made file. It is 400 pages.

Tools this post uses: