DuckConvert Mascot
DuckConvert
Blog

Compress PDF

Shrink a scanned document by re-encoding the images inside it, in your browser. Text stays text.

Compress PDF

Re-encodes the photographs and scans inside the document. Text and vector drawings are left untouched, because they are already small.

Where the size in a PDF actually is

A PDF is a bag of objects: page trees, font programs, drawing instructions, and image streams. Everything except the images is already tiny and already compressed. A forty-page report exported from a word processor might be a few hundred kilobytes in total, and no amount of squeezing will make it meaningfully less.

A four-page scan of the same document can be twenty megabytes, because every page is a photograph at 300 dots per inch. That is where the entire opportunity lives, which is why this tool works on images and leaves everything else alone.

Two levers, both applied

The first is the JPEG quality the image is re-encoded at. The second, and usually the bigger one, is resolution: a 300 DPI scan is about 2550 pixels across an A4 page, and reading it on a screen or printing it at 150 DPI needs half of that. Halving both edges removes three quarters of the pixels before the encoder has done anything.

What it deliberately does not do

It does not rasterise your pages. Some compressors reach their headline numbers by rendering each page to a single flat image and wrapping those in a new PDF. The file does get smaller, and your text stops being text: no selecting, no searching, no copying, and a document that looks soft at any zoom. Here the text objects are carried across untouched and only image streams are replaced.

It also refuses to hand back something worse than it was given. If the images were already compressed at least as hard as you asked for, rebuilding the document would cost more bytes than it saved, so you get the original back exactly as it was and are told why. A compressor that returns a bigger file and calls it a success is not compressing.

Compress PDF FAQ

Is my PDF uploaded to be compressed?

No. The document is parsed, its embedded images decoded and re-encoded, and the file rebuilt, all by code running on your own device. The PDFs people most want to shrink are the ones they are about to email — a signed contract, a bank statement, a medical scan — and those are exactly the files that should not be passing through a stranger's server on the way.

Why did my PDF barely get smaller, or not change at all?

Because it had nothing worth re-encoding. A PDF exported from a word processor is text and vector drawing instructions, which are already about as compact as they get — there is no equivalent of turning down the quality on a paragraph. All the room in a large PDF is almost always in embedded photographs and scans. When there are none, or when they are already compressed at least as hard as the setting you chose, you get your original file back byte for byte rather than a rebuilt one that happens to be slightly bigger.

Which images does it actually re-encode?

JPEGs, which is what a scanner or a camera puts in a PDF and where essentially all the size is. Their stream inside the document is a JPEG file, so it can be handed straight to your browser's decoder and back. Images stored in other ways are left exactly as they are: some need decoders no browser has, and others are raw pixel data whose meaning depends on the document's own colour setup, where a wrong guess produces a wrongly-coloured page rather than a smaller one.

Will the text still be selectable afterwards?

Yes. Only the image streams are replaced; the text, fonts, links, bookmarks and page structure are carried across unchanged. This is not a re-print of the document into pictures of pages, which is how some compressors get their impressive numbers — and how they quietly destroy every piece of searchable text in the file.

Is anything lost?

Yes, and it is worth being clear about it: JPEG re-encoding is lossy, and the images are also scaled down when they are larger than the setting allows. A 300 DPI scan reduced for on-screen reading will not print quite as crisply as the original. That is the trade you are making, which is why there are three settings and why the result tells you exactly how much came off rather than just claiming success.