Bad PDF compression is easy to spot: fuzzy letters, gray smudges where signatures were, tables you can no longer read. Good compression targets what actually makes PDFs big — and in almost every oversized document, that's the images, not the text.
Where the megabytes live
PDF text is vector data: mathematically described letter shapes, plus fonts. Even long documents cost only kilobytes of text. Scanned pages, photos, and screenshots are bitmaps — and one carelessly scanned page can outweigh a hundred pages of text. So the right order of operations is:
- Recompress the images, keep everything else. Re-encode embedded photos/scans at a saner JPEG quality (and, if needed, slightly smaller dimensions). Text, fonts and layout are untouched, so sharpness is preserved where it matters most. This is the default mode of our PDF Target Size tool.
- Drop dead weight. Editing metadata (XMP) can carry surprising bulk in some generators; removing it costs nothing visually.
- Rasterize only as a last resort — and knowingly. Converting whole pages to images reaches very small targets, but text stops being selectable and zooming gets blocky. For documents that are already scans, this costs almost nothing; for born-digital text documents it's a real trade, which is why our tool discloses it and asks before using it.
Grayscale: the honest extra lever
Color triples the information in a scan. If the portal doesn't need color (most text documents don't), grayscale conversion buys a large reduction with zero readability cost.
What a trustworthy tool should tell you
- The actual output size, verified against your target — not “compressed!” with no numbers.
- Which method was used, especially if pages were rasterized.
- An honest failure when a target isn't reachable — a 40 MB photo album will not become 500 kB and still look like anything.