Before compressing anything, find out what you are actually dealing with — because the effective fix is completely different depending on the cause, and most people apply the wrong one.
A hundred-page text document is typically a few hundred kilobytes. A three-page scanned contract can be twenty megabytes. Page count is almost never the problem. Images almost always are.
Work out where the weight is
The quickest diagnosis: does the document contain scanned pages or photographs? If yes, that is your answer and you can stop investigating.
If not, the likely causes in order:
Images placed at far higher resolution than they are displayed. A 4000-pixel logo appearing two centimetres wide carries roughly forty times the data it needs. Extremely common in documents built from a template.
Fonts embedded without subsetting. A full font family can run to several megabytes. Subsetting embeds only the glyphs actually used and typically cuts that by 90%.
Accumulated revision history. PDFs can be saved incrementally, keeping every earlier version inside the same file. A document edited fifty times may contain fifty versions.
Complex vector graphics. Detailed maps, CAD exports, and heavy charts carry a great deal of path data.
For a proper look, Acrobat Pro has an "Audit space usage" report that itemises exactly where the bytes are. qpdf --json or mutool info give similar information free.
What actually shrinks a file
Downsample the images. The single most effective change and the one to try first. A 600 DPI scan reduced to 150 DPI is a quarter of the pixels and looks identical on screen. Resolution is the real lever — far more so than JPEG quality settings, which people reach for first and which cost visible quality much faster.
Re-encode the images. Stronger JPEG compression on photographic content. Effective, with diminishing returns and rising artefacts as you push it.
Strip redundancy. Unused objects, duplicate resources, old revisions, and orphaned metadata all go without touching quality at all. Free savings.
Subset the fonts, if they were embedded whole.
Flatten pages to images. The last resort. It shrinks almost anything and destroys selectable text, search, links, and accessibility. Only for a document that is purely visual, or when nothing else has worked.
Doing it
In the browser. Our Compress PDF tool runs entirely on your machine — no upload, no account, no limits. Drop the file in, choose Light, Recommended, or Strong, and download. It recompresses images while leaving text selectable, which on image-heavy documents commonly saves 50–90%. There is also an opt-in Maximum compression mode that flattens pages to images; it will tell you when it has done that, because you should know.
Ghostscript, free and local, is the most capable command-line option:
gs -sDEVICE=pdfwrite -dPDFSETTINGS=/ebook -dNOPAUSE -dQUIET -dBATCH \
-sOutputFile=out.pdf in.pdf
/screen is smallest, /ebook a good balance, /printer and /prepress preserve more. Excellent for batches.
Print to PDF. Sometimes helps by discarding accumulated cruft, sometimes makes things worse by re-encoding. Free to try. It also strips links, bookmarks, and accessibility tags, so treat it as a last resort rather than a technique.
Preview on macOS. File → Export with the "Reduce File Size" Quartz filter. Convenient, and fairly aggressive — check the result.
Online compressors — Smallpdf, iLovePDF, Adobe — work well and upload your file. For anything sensitive, use one of the local options above.
Fix it at the source instead
If you are producing the PDF yourself, this beats compressing afterwards every time, because you never create the waste.
- Resize images before inserting them. Do not drop a 4000-pixel photo into a document where it appears as a thumbnail.
- Compress images first with an image tool.
- Export with the right preset. Word, InDesign, and most tools offer "smallest file size" or "web" options that downsample automatically.
- Scan sensibly. 300 DPI greyscale for text documents, not 600 DPI colour. This one decision routinely accounts for a tenfold difference. See compressing a scanned PDF.
When not to compress
Print production. Commercial printers need full resolution — usually 300 DPI at final size — and a compressed file will print visibly soft. Send the high-resolution version and let them handle it.
Archival and legal records. Compression is lossy; the archive copy should be the original. Compress a working copy if you need to send one.
Documents where fine detail carries meaning — technical drawings, medical imaging, maps, signatures being examined for authenticity.
Files that are already small. Squeezing a 400 KB document is effort spent for nothing.
If it will not shrink
Usually because it was already optimised. A PDF exported from a modern application is often close to minimal, and pushing further trades real quality for a few percent. When that happens, the better answers are to split the document and send only the relevant pages, or to share a link rather than an attachment.
And check what you are aiming for before you start. Gmail and most providers cap attachments around 25 MB, many corporate mail servers at 10 MB, some portals at 5 MB. Knowing the target stops you compressing far harder than necessary. See compressing a PDF for email attachments and why is my PDF so large?.
