Compress PDF

A PDF is big because of its pictures

Text and vector shapes in a PDF are drawing instructions, and instructions are tiny. A forty page contract with embedded fonts is perhaps a megabyte. What makes a document enormous is images: a scan is a photograph of every page, and phone photos dropped into a report are twelve megapixels each.

So compressing a PDF means finding those images and writing them back smaller. Everything else is rounding error, which is why "compressors" that only re-deflate the file's structure report three percent and leave people confused. What a PDF actually is covers the rest of what is in there.

The number nobody else works out

"Reduce images to 150 dpi" only means something if you know what dpi each image is at, and an image has no resolution of its own: a 4,000 pixel picture is 480 dots per inch across an A4 page and 4,000 across a postage stamp. Only the first is worth touching.

So this reads each page's content stream and follows the transformation matrix to find the size every picture is actually drawn at, then reduces only the ones above the target. A photograph spread across a page is shrunk; the small logo in the corner, already sharp and already small, is left alone. That is what the expensive desktop tools do, and it is the difference between a document that still reads well and one where the letterhead has gone fuzzy. What dpi actually means has the arithmetic.

Three targets, because there are three reasons

Printer, 300 dpi. The standard for anything that will come out on paper. A scan compressed to this still prints properly.

Screen, 150 dpi. Sharp on any display, and usually a half to a tenth of the original. The right default for a document people will read rather than print.

Email, 96 dpi. For when a form or a mailbox has given you a hard number to get under. It is visibly softer if you zoom in, and it is the one that gets a 40 MB scan attached to a message.

What it deliberately will not touch

Anything it cannot decode without guessing, because a wrong decode is a corrupted document rather than a smaller one. That means JPEG 2000, CCITT fax and JBIG2 images, which are the formats a scanner uses for black and white pages and are already far smaller than anything this could write; indexed and separation colour spaces; and one-bit stencil masks, where re-encoding as a photograph would blur the very edge the mask exists to define.

The row for each file says how many pictures were rewritten and how many were left alone, so a document that comes back the same size tells you why rather than leaving you to try every setting. And it never hands back a file larger than the one it was given: if nothing could be improved, you get the original untouched.

It is not lossless, and nothing useful is

Reducing an image's resolution and re-compressing it throws detail away. That is the mechanism, not a side effect, and it is one-way: the detail does not come back when you change your mind. Keep the original, and compress a copy. Does converting to PDF lose quality goes through what is and is not recoverable.

The one case where compression is genuinely free is a document whose images were stored uncompressed or losslessly when they should have been JPEGs, which is common in exports from older software. There the file can drop by ninety percent with no visible change at all.

Many at once, and none of them uploaded

Drop a folder's worth. Documents are done one after another, each with its own before and after, and Save all puts them in one ZIP. There is no cap on the number or the size, because there is no server to pay for.

That last part matters more here than on most pages. The PDFs people compress are the ones an upload form just refused: mortgage applications, visa packs, medical letters, scans of passports. Those are exactly the files that should not be handed to a stranger's machine to be made smaller. Load this page, go offline, and compress one anyway if you want to see for yourself.

Nearby

If the document is too big because it holds too many pages, split PDF is the better answer. Merge PDF is the other direction, and compress PNG is the page for pictures that are too big before they ever reach a document.