How to compress a PDF (and why some barely shrink)

A PDF is a container, not a format

Inside a PDF there are really only a few kinds of thing: text with the fonts needed to draw it, vector art (lines, shapes, charts), and raster images. Text and vector art are instructions — "draw this letter here, in this typeface" — and they are tiny. A page of dense body text is a few kilobytes.

Raster images are pixels, and pixels are heavy. A single photograph scanned at 300 DPI covering an A4 page is roughly 8.7 million pixels. Even well compressed, that is hundreds of kilobytes for one page.

So when you look at a 40 MB PDF, the useful question is not "how do I compress a PDF" but "what is the 40 MB actually made of". Almost always the answer is images, and almost always they are bigger than they need to be.

Why scans shrink and text documents do not

A scanner does not know the difference between a letter and a smudge. It photographs the page, so every page becomes one large image — usually at 300 or 600 DPI, because that is what the scanner defaults to, not because anyone needs it. On a screen you will never see past about 150 DPI, so more than half the pixels in that file are being stored for nobody.

That is exactly the space compression reclaims. Downsampling those images to a sensible resolution and re-encoding them is where the 70–90% comes from, and on a scan you genuinely cannot see the difference on screen.

A text document has none of that slack. Its fonts are already subset and deflated, its text is already compressed, and there are no oversized images to trim. A rewrite recovers a few percent from repacking the internal structure, and that is honestly all there is. If a tool promises you 80% on a text PDF, it is either about to rasterize your text — destroying your ability to select, search or copy it — or it is guessing.

Choosing a level you will not regret

The only real question is where the document is going to be read. Balanced targets 150 DPI: crisp on any screen, and still perfectly acceptable on paper. It is the right answer for almost everything — email attachments, uploads to a portal, documents you archive and read.

Choose high quality when the file will actually be printed at size, or when someone will zoom into a diagram. It keeps images at 200 DPI, so the saving is smaller but nothing softens under magnification.

Choose maximum compression when a hard size limit is the whole problem — a form that refuses anything over 5 MB, for instance. At 96 DPI photographs will visibly soften if you print them, but on screen they stay readable, and a document that fits is worth more than one that does not.

Whichever you pick, the text stays text. Fonts are subset to the glyphs your document uses, which makes the file smaller without touching a single character on the page.

What compression will not do

It will not open a password-protected file. An encrypted PDF cannot be rewritten without its password — we detect that before uploading anything and say so, rather than failing halfway through. Remove the protection in your PDF reader first.

It will not reorder, remove or reflow your pages. The page count of the finished file is checked against the original before it is handed back; if they differ, the job fails rather than giving you a document that quietly lost something.

And it will not make a file bigger. If a rewrite comes out larger than what you started with — which happens with already-optimized PDFs — we keep your original bytes and tell you it was already as small as it gets. A compressor that returns a bigger file has not compressed anything.