How PDF Compression Actually Works: Downsampling, Subsetting, and More

Images Are Almost Always the Real Problem

A 50-page PDF of pure text rarely exceeds 200 kilobytes. Add a single unoptimized photograph, and that same file can jump to 15 megabytes or more. Embedded images are, by a wide margin, the dominant factor in most oversized PDFs — which is exactly why image handling is where the majority of compression work happens.

Image Downsampling (Lossy)

A document scanned or created for professional printing often embeds images at 300 DPI or higher — far more pixel data than a screen needs to display the same image clearly. Downsampling reduces that pixel density to a lower target (150 DPI for screen viewing is a common recommendation), discarding pixels that a normal-zoom viewer would never notice missing. This is the one genuinely lossy step in PDF compression — the discarded pixel data cannot be recovered — but at reasonable target resolutions, the visual difference is imperceptible at normal viewing distances.

Font Subsetting (Lossless)

Embedding a font in a PDF often includes the entire font file — every character it defines, even if your document only actually uses a fraction of them. Font subsetting keeps only the specific characters your document actually contains, discarding the rest. This is completely lossless: your text renders identically, just with less unused font data along for the ride.

Removing Redundant Data (Lossless)

PDFs edited multiple times can accumulate duplicate objects, unused metadata, and leftover revision history that no longer serves any purpose. Cleaning this out, and applying standard lossless compression (Flate/ZIP) to text and vector content streams, further reduces size without touching anything visible.

Why This Isn't the Same as Just Zipping the File

Wrapping an already-structured PDF in a generic ZIP archive typically achieves very little — a 100MB PDF zipped might only shrink by a few percent, since PDF's internal structure is already partially compressed and ZIP doesn't understand PDF-specific redundancies like oversized embedded images. Proper PDF compression works inside the file's actual structure, targeting the specific things — images, fonts, redundant objects — that are actually inflating its size.