You compressed it and it barely moved
A scanned PDF is the file people most often need to shrink and the one that most often refuses. You run it through a compressor, it drops from 18 MB to 16 MB, and the upload form still says no. The reason is that a scan is not a document in the usual sense: it is a stack of photographs of paper, one per page, and photographs only get smaller when you throw pixels away. This guide explains what a scanner actually writes into the file, why generic compression does nothing, and the specific changes that do work — with numbers, so you can predict the result before you start.
What a scanner puts in a PDF
A flatbed or sheet-fed scanner set to 300 dpi turns an A4 page into an image of about 2,480 × 3,508 pixels — 8.7 million of them. In colour, uncompressed, that is 26 MB per page. The scanner software then encodes the image, and the choice it makes decides everything that follows:
| Encoding the scanner used | Per A4 page at 300 dpi | How to tell | What compression can gain |
|---|---|---|---|
| JPEG, high quality (most home scanners, phone scan apps) | 1.5–4 MB | Photo-like, slight smudging around letters at high zoom | A lot — quality is set far higher than a document needs |
| JPEG, medium quality | 400–900 KB | Visible blockiness at 300% zoom | Some — mainly by lowering resolution |
| Lossless (Flate/ZIP or uncompressed) — office copiers, 'archive' modes | 5–25 MB | File size wildly out of proportion to page count | Enormous — 10–30× by switching to JPEG |
| CCITT G4 / JBIG2 (black-and-white fax-style) | 30–80 KB | Pure black and white, no greys | Almost none — already the most compact form |
Generic PDF compressors that work on the file structure — removing unused objects, deduplicating fonts — cannot touch any of this, because there is no structure to trim: the file is 99% image data. That is why a tool that shrinks a Word-exported PDF by half does nothing for a scan.
The three things that actually shrink a scan
Size of an image page ≈ width × height × bits per pixel × JPEG factor. Each of the three changes below attacks one term. Combined, they routinely turn a 20 MB scan into 1 MB.
- 1. Lower the resolution. Halving dpi cuts pixels by four. 300 dpi is for OCR and reprinting; 150 dpi is perfectly readable on screen and prints acceptably; 100–120 dpi is fine for a reference copy. This is the single biggest lever — a 300→150 dpi change alone is a 4× reduction.
- 2. Drop colour. A black-on-white page carries no useful colour. Greyscale stores one channel instead of three, and JPEG compresses greyscale text more efficiently. Expect 40–60% off at the same resolution. Keep colour only for pages where colour carries meaning — stamps, highlighted clauses, photographs of ID cards.
- 3. Lower JPEG quality. Scanner defaults sit around quality 0.9–0.95, which is wasted on paper. Quality 0.7 is visually identical for text at normal zoom; 0.5–0.6 is where softening becomes noticeable. Going from 0.92 to 0.7 typically halves the size.
On this site, Compress PDF applies levers 1 and 3 (it re-renders each page at the dpi and quality you pick), and Grayscale PDF applies lever 2. Run Grayscale first, then Compress, and you get all three.
Worked example: a 20-page colour scan
| Step | Setting | Size | Reduction |
|---|---|---|---|
| Original from scanner | 300 dpi, colour, JPEG ~0.92 | 22 MB | — |
| Grayscale PDF | 300 dpi, grey | 11 MB | 2× |
| Compress PDF | 150 dpi, quality 0.75 | 1.6 MB | 14× |
| Compress PDF (tighter) | 110 dpi, quality 0.65 | 0.8 MB | 27× |
The 1.6 MB version is indistinguishable from the original at normal reading size and still prints cleanly. The 0.8 MB version is for a 1 MB portal cap; small print goes slightly soft but stays legible.
Things that make scans bigger than they need to be
- Scanning in colour by habit. Set the scanner to greyscale (or black-and-white for pure text) before scanning; it beats fixing it afterwards.
- 600 dpi. Nothing on paper needs it except tiny type destined for OCR. It quadruples 300 dpi.
- Wide margins and blank backs. Pixels of white paper still cost bytes. Crop PDF trims margins; Delete PDF Pages removes blank reverse sides from duplex scans.
- Phone scan apps' 'enhance' filters add contrast noise that JPEG compresses badly. Turn enhancement off and let the compressor do its job.
- An OCR text layer adds only a few KB per page — keep it, but note that rasterized compression discards it (see below).
What you lose, and how to keep it
Re-rendering pages removes any invisible OCR text layer the scanner added, so the compressed file is no longer searchable. If you need both small and searchable, compress first, then run OCR on the result in Acrobat, macOS Preview, or Google Drive — OCR works fine at 150 dpi. Form fields and annotations are also flattened into the image; on a scan that is rarely a concern, since a scan has none to begin with.