compress Compress PDF star

How PDF Compression Actually Works

Open almost any bloated PDF and the weight is concentrated in one place: embedded raster images, not the text. A scanned page is a full-resolution photograph of paper — often 300 dots per inch (dpi) or more — and a handful of those photographs can outweigh hundreds of pages of pure text by an order of magnitude. Compression works by attacking exactly that: it downsamples each embedded image to a lower resolution (a common target is 300dpi → 150dpi, still sharp on a screen and adequate for most printing) and re-encodes it as JPEG at a chosen quality setting, discarding the fine detail a screen or a standard printer cannot show anyway.

The two levers — resolution and JPEG quality — trade off independently. Halving the resolution alone cuts roughly to a quarter of the pixel data (both dimensions shrink), while lowering JPEG quality reduces the file further by throwing away high-frequency detail through more aggressive quantization. A typical "medium" preset combines a moderate downsample with 70-80% JPEG quality, which on a photo-heavy scan often produces a 60-90% size reduction with no visible loss on screen.

Critically, this only touches raster content. Native text, vector line art and embedded fonts are stored as instructions (draw this glyph, this shape) rather than pixels, so they compress losslessly or are left untouched entirely — the words stay just as sharp and just as selectable after compression as before. That is also why compressing a text-only, born-digital PDF barely shrinks it at all: there is no image weight to remove, so the tool is squeezing a file that was never carrying the excess in the first place.

What to Keep in Mind

• A PDF that is already a photograph of every page (a flat scan saved through a rasterizing print driver) benefits the most — it is nothing but the exact content this method targets.
• Higher quality settings preserve more detail but shrink the file less; there is no setting that improves both simultaneously, because the trade-off is intrinsic to how JPEG and downsampling work, not a limitation of any particular tool.
• If a page mixes a background photo with overlaid text (as many scanned forms and slides do), only the photo layer is degraded — but because the two are flattened into one image on many scans, text baked into the picture will soften along with it.
• Repeated compression compounds loss: compressing an already-compressed PDF a second time re-quantizes JPEG data that has already lost information, so always compress from the highest-quality original you still have, not from a previously compressed copy.