How do you actually reduce a PDF’s file size?
Almost all PDF bulk is embedded images, so shrinking their pixel dimensions is the fix that works. Removing pages, fonts or metadata saves far less than people expect, and re-compressing already-compressed scans usually degrades quality for almost nothing.
Updated 2026-08-22
Find out what is actually heavy
A PDF is a container, and the useful question is which of its contents is large. In practice there are only three candidates: embedded images, embedded fonts, and everything else. Everything else is almost never the answer — page text, vector drawings and the document structure are typically a few kilobytes a page even in long documents.
Embedded images are the answer perhaps nine times in ten. A single photograph from a modern phone is three to eight megabytes; twenty of them assembled into a document are what turn a PDF into a file that will not send. Scans are the same thing under a different name, because a scanner produces photographs of paper.
Fonts are the occasional second cause. A fully embedded font family can add several hundred kilobytes, and a document that embeds four weights of two families carries that whether or not the pages use them. This matters for a five-page document that is inexplicably two megabytes with no pictures in it.
The quick diagnostic without any tool: how many pages, and do they contain photographs or scans? A forty-page text document at eight megabytes has a font or image problem. A six-page scan at eight megabytes is behaving exactly as expected and the pages are the problem.
Resize the images, do not re-compress them
This is the single highest-value action and it is consistently the one people skip, because "compress" sounds like the operation and "resize" does not. File size scales with pixel count. An image 4000 pixels wide contains four times the data of the same image at 2000 pixels wide, and if it is being printed on A4 or read on a screen, 2000 is already more than anyone will see.
Compare that with quality tuning. Dropping JPEG quality from 90 to 75 might save a third and introduces visible artefacts. Halving the dimensions saves three quarters and, at normal viewing size, looks identical. The two are not close, and only one of them degrades what you can see at the size you actually view it.
The practical route is to fix this before assembly rather than after. Resize the photographs, then build the PDF from them. Once they are embedded, most compressors can only re-encode what is there — which means decoding an already-lossy image and encoding it again, compounding the damage for a saving that is often disappointing.
If the document already exists and cannot be rebuilt, accept a smaller win, or extract the pages that matter and leave the rest behind.
Why compressing a scan barely helps
A scanned PDF is a stack of JPEGs. Those JPEGs were compressed when the scanner made them, and compression only works once — the redundancy has already been removed. Running the document through a compressor asks an encoder to find savings in data from which savings have already been extracted.
What it usually does instead is re-encode: decode each image and compress it again, more aggressively. That produces a smaller file and a worse-looking one, and because JPEG artefacts themselves look like detail to the next encoder, the ratio is poor. This is why "compressed" scans so often arrive blotchy and only 20 percent smaller.
The exception is a scan saved at a genuinely excessive resolution — 600 dpi for a document nobody will print. Downsampling that to 200 or 300 dpi is a resize, not a re-compression, and it works for exactly the reason in the previous section.
Things that save less than you think
Removing metadata. Worth doing for privacy, but the document information dictionary is measured in bytes. It will not move a file that is measured in megabytes.
Deleting a few pages. This helps in proportion to what those pages contained. Dropping two text pages from a forty-page report saves almost nothing; dropping two scanned pages from a ten-page scan saves a fifth. The saving comes from the images that left with them, not from the page count.
Flattening or printing to PDF. This can help when a document carries form fields, annotations or layers, and it can easily hurt otherwise, because some "print to PDF" paths rasterise the whole page — turning crisp vector text into an image and making the file larger as well as unsearchable.
Zipping it. A PDF is already compressed internally, so a zip typically saves single-digit percentages and adds a step for the recipient. It is worth doing only to bundle several files together, not to shrink one.
The order to try things in
First, ask whether the recipient needs the whole document. Extracting the four relevant pages is instant, lossless, and often the entire problem solved — and a smaller document is a better one to receive regardless of file size.
Second, if you still have the source images, resize them and rebuild. This is the biggest genuine win available and it costs no visible quality at normal viewing sizes.
Third, if the document is a scan you cannot rebuild, check the resolution. Downsampling from an excessive dpi is worthwhile; re-compressing an already-reasonable scan is not, and will cost you legibility for a modest saving.
Fourth, if it is a text document that is mysteriously heavy, the cause is font embedding and the fix belongs in whatever produced it — subset the fonts on export rather than trying to repair the finished file.
And throughout: none of this requires uploading the document to anyone. Every step above can be done on your own machine, which matters more than usual here, because the documents people most want to shrink are the ones they least want to hand over — contracts, scans of identity documents, medical records, tenders.