Compress PDF

If you are here because a form will not take anything over two megabytes, tick the size box and type the number. The best-looking setting that fits is the one you get, and if the file cannot go that small you are told the smallest it reaches rather than handed one that still will not upload. Otherwise pick a quality. A scanned PDF is almost entirely photographs, so making one smaller means re-compressing those and leaving everything else alone, and a file with nothing worth compressing is told so rather than handed back with a new name.

1 · Your PDFs

2 · How hard to squeeze

Quality

Balanced suits almost everything. The FAQ below says exactly what each one changes.

3 · The result

Add a PDF to begin. Several at once is fine.

Frequently asked questions

Can it really get any file under the size I ask for?

Often, and not always, and it tells you which before you download anything. The floor is about fifteen to twenty kilobytes a page: at the smallest setting each page becomes a picture around 650 pixels on its longest side, and no amount of asking makes it smaller than that. So a 2 MB target holds for roughly a hundred pages of scan and not for four hundred. Measured on files built for the purpose: a 93 MB, 50-page scan reaches 1.03 MB in 19 seconds, and a 90 MB, 220-page one stops at 3.39 MB and says so rather than handing you something that will still be rejected.

Is my document uploaded?

No. Your browser reads the PDF, decodes each photograph, encodes it again and writes a new file, all locally. The site’s security policy sets connect-src ‘none’, so the browser blocks network requests from these pages, meaning the document could not be sent even if the code tried.

Why did nothing happen to my file?

Because there was nothing to compress. A PDF of text is already small: the letters are instructions, not pictures. The size of a PDF is almost entirely the photographs inside it, so a document without any cannot be made meaningfully smaller by any tool, whatever it claims.

Which images get re-compressed?

The ones already stored as photographs. Those are scans and camera pictures, which is exactly what re-encoding helps. Images stored losslessly are screenshots, diagrams and line art, where converting to a photograph format makes a bigger file that looks worse, so they are left alone. The panel reports how many of each it found.

What about pictures with transparency?

Left alone. The transparency is stored in a separate object whose size has to match the picture, and changing one without the other gives the image someone else’s holes. Getting that right is worth doing, but not worth guessing at.

What is the difference between the three settings?

Each one changes two things at once: how hard each photograph is compressed, and how large it is allowed to stay. Smallest file compresses hard and caps pictures at 1200 pixels on the long side, which is right for a scan you only need to read. Balanced compresses moderately and caps at 1600, which is right for almost everything. Best looking compresses gently and caps at 2000, which is right when the photographs matter and the file only needs to be somewhat smaller. On a fifty-page scan of 2400 pixel photographs the three save about 80, 62 and 24 per cent.

Why is there no percentage to set?

Because the number was never the decision. Nobody opens a document and thinks "seventy-one per cent"; they think "this needs to be small enough to email" or "these photographs have to stay sharp". A slider makes you translate your intent into a number and guess whether you got it right, and the result only appears after the work is done. Three named outcomes ask the question you were already answering.

Does capping the size help more than quality?

Usually, yes, which is why every setting does both. A phone photograph dropped into a document is often four times larger than it will ever be displayed, and halving its width removes three quarters of the pixels. Quality reduces the cost of each pixel; the cap reduces how many there are.