Compress a scanned PDF

Why this works

Our scan-optimised compression detects image-only pages, applies smart downsampling and re-encodes with JBIG2 / JPEG2000 where supported. Typical reductions are 70–95% with text still clearly readable.

Scanned PDFs are size hogs because of how they\'re built. A born-digital PDF stores text as text — the actual character codes, taking roughly 1–2 KB per page. A scanned PDF stores each page as a high-resolution photograph of paper, taking 3–6 MB per page in colour at typical scanner settings. A 20-page scanned contract that exists as text would weigh 50 KB; as a scan it weighs 80–120 MB. Same content, 2,000× the file size.

Default scanner settings are the root cause. Office multifunction scanners typically default to 300 DPI in colour, which is appropriate for photographs but enormously overkill for documents that contain only text. Phone scanner apps (Adobe Scan, CamScanner, Microsoft Lens) default to similar settings or even higher to maximise OCR accuracy. Copy-shop self-service scanners are often set to maximum quality by default because operators can\'t predict whether customers want archival photos or text documents.

What the scan-optimised compression actually does, in order: detect image-only pages (no embedded text layer) versus mixed pages (some text, some image) and apply different treatment to each; downsample images from typical scanner output (300 DPI) to 150 DPI — imperceptible on screens, modest impact when printed; convert colour scans to grayscale when colour adds no information (most office documents); apply JBIG2 compression for monochrome pages — a specialised compression algorithm designed for scanned text that achieves 50–80% better ratios than generic JPEG; apply JPEG2000 compression for greyscale and colour pages where supported; strip duplicate page resources (e.g. recurring scanner-bed shadow patterns across pages).

A practical sequence for getting scanned PDFs to manageable sizes. First, run OCR if you haven\'t already — OCR adds an invisible text layer but doesn\'t bloat the file much (typically <100 KB total per page), and once you have OCR\'d text, downstream compression can be more aggressive on the image layer without losing searchability. Then run scan-optimised compression with the Recommended setting — typical reduction is 60–80% with all text readable. If you need to hit a specific size cap (1 MB for visa portals, 10 MB for email), escalate to Aggressive or also convert to grayscale.

Where compression hits limits. Photographs of architectural drawings, blueprints, or detailed diagrams can\'t compress as aggressively because every pixel encodes information — expect 30–50% reduction rather than 80%. Documents with red-ink notations, highlighter marks, or coloured stamps lose information when converted to grayscale; keep colour mode for these. Hand-annotated documents with complex sketches may compress poorly if the annotations are essential to the document\'s meaning.

For very long scanned documents (50+ pages), consider splitting first. A 100-page scan compressed and sent as one file is often unwieldy regardless of the byte count; splitting into logical sections (chapters, sections, monthly statements) makes the recipient\'s life easier and lets you re-send only the affected section if a correction is needed later.

How it works