PDF compression, explained honestly

What is actually inside a PDF

A PDF is a container holding a few kinds of cargo, and each kind responds differently to compression:

• Images — photos, scans, screenshots. In a scanned document, images are essentially the whole file. This is where the megabytes live.
• Text and vector graphics — real text, lines, tables. Already tiny and already compressed; a 300-page text-only novel can be under 2 MB.
• Fonts — embedded typefaces, often a few hundred kilobytes each. A file using many fonts (or several full CJK fonts) carries real weight here.
• Baggage — thumbnails, deleted-but-retained objects, duplicate resources left behind by whatever software built the file.

This explains the mystery immediately: a scanned PDF shrinks dramatically because it is made of photographs, and photographs can be resampled and re-encoded. A digitally-created, text-only PDF barely shrinks because there is almost nothing left to squeeze.

What a compressor actually does

Three things, in roughly descending order of impact:

1. Downsampling images. A page scanned at 600 DPI holds four times the pixels of the same page at 300 DPI, and sixteen times 150 DPI. For on-screen reading, 150 DPI is generally indistinguishable; for printing, 300 DPI is the standard. Dropping excess resolution is where the big wins come from.
2. Re-encoding images at a stronger JPEG quality setting. Going from quality 95 to quality 75 often halves an image's bytes with changes you need a loupe to see.
3. Housekeeping — stripping unused objects, deduplicating resources, rebuilding the file structure. Modest gains, but free: no visual change at all.

Choosing a quality level without guessing

Think about the document's future, not its present:

• It will only ever be read on screens (email attachments, portal uploads, archives): compress hard. 150 DPI, moderate JPEG quality. This is the right call for most paperwork.
• Someone might print it: keep 300 DPI. The file stays bigger; that is the honest cost of print quality.
• It contains fine detail that is the point — engineering drawings, small-print terms, faint stamps or seals: compress gently or not at all, and check the result page by page.

One warning that saves real grief: compression is cumulative and one-way. Re-compressing an already-compressed scan stacks JPEG artefacts on JPEG artefacts, and text edges grow fuzzy halos. Keep the original file; compress copies.

When compression “doesn't work”

If the file barely shrinks, one of these is true: it is text-and-vector with nothing to squeeze (already small — check whether it actually needs compressing); its images were already downsampled by a previous pass; or the weight is in embedded fonts, which most compressors rightly refuse to butcher. And if a scanned file refuses to shrink, it may be encoded as PNG-style lossless images — converting those to JPEG is exactly what a compressor's stronger settings do, so try the next level down.

The 40 MB question

Email providers commonly cap attachments around 20–25 MB, and many government and job portals cap uploads far lower. That, in practice, is why people compress PDFs at all. The workflow that works: compress a copy at the screen-reading level with a tool like our compressor (which does the work in your browser — the file never leaves your machine), check the pages you care about at 100% zoom, and only reach for the stronger setting if the target size demands it. Compress with a purpose, keep the original, and the process holds no surprises.

Keep reading