Compress a Scanned PDF Without Making It Unreadable

Most advice about compressing scans tells you how much smaller you can make the file. Far less of it tells you when you have gone too far — and that is the question that actually matters when the document is a rental application, a visa submission or a signed contract.

A scan that is 300KB and unreadable is worse than one that is 3MB and fine. The mechanics of compressing a scan are covered in how to compress a scanned PDF. This guide is about the threshold: where legibility breaks, and how to verify your file before you hand it over.

Where the readability floor sits

Resolution is the setting that determines whether a scan can be read. Everything else is secondary.

The practical rule: 150 DPI is the floor for normal documents, 200 DPI is the floor for difficult ones.

At 150 DPI, a standard 11pt character is about 23 pixels tall. That is enough to render letter shapes cleanly. At 96 DPI the same character is 15 pixels, and the difference between an 8 and a B, or a 5 and an S, starts depending on two or three pixels — which is exactly where misreadings happen on account numbers and reference codes.

What pushes the threshold up

Go to 200 DPI rather than 150 when the document has any of these:

• Small print. Terms and conditions, footnotes, legal boilerplate — anything below about 9pt.
• Handwriting. Signatures, handwritten form entries, annotations. Handwriting has thinner, more variable strokes than print and degrades faster.
• Reference numbers. Policy numbers, account codes, NI numbers. A single misread character makes the document useless.
• Faint originals. Carbon copies, faded receipts, thermal paper, anything photocopied several times.
• OCR afterwards. Text recognition is unreliable below 200 DPI. If you need the document searchable, 200 DPI is not optional — see make a scanned PDF searchable.
• The recipient will print it. Office printers run at 300–600 DPI. A 150 DPI scan prints acceptably; a 96 DPI one looks obviously degraded on paper.

The settings that do not cost readability

Two of the three compression levers are nearly free on a typical document scan.

Greyscale is free on text documents

Scanners default to colour. Most documents are black text on white paper. Converting to greyscale removes two of three colour channels for content that only ever used one, typically cutting half to two-thirds of what remains after downsampling.

It costs nothing in readability because there was no colour information worth keeping. Keep colour only where it carries meaning — highlighting, coloured charts, official stamps, or a blue-ink signature where that matters evidentially.

JPEG quality has a lot of headroom

75–80% quality is indistinguishable from the original on a document. The damage starts below about 60%, and it appears first as ringing artefacts around letter edges — soft grey halos that make characters look fuzzy.

The order matters: drop the resolution first, convert to greyscale second, and only touch JPEG quality if you still need more. Resolution gives you far more saving per unit of damage.

Black-and-white mode is the exception

Black-and-white (1-bit) mode is dramatically smaller — a clean printed page can be under 50KB — and for crisp typed text on white paper it is both the smallest and the sharpest option, because it eliminates JPEG artefacts entirely.

But it is unforgiving. Anything with shading, a photograph, a greyscale logo, a photo ID or a faint pencil annotation renders as harsh dithering or disappears altogether. Use it only on clean, high-contrast printed text, and check the result carefully.

Compressing an existing scan

Our PDF compressor runs entirely in your browser, which matters for this category of document specifically — passports, payslips, medical letters and signed agreements should not be uploaded to an unknown server.

1. Drop the PDF in.
2. Choose Downsample pages to images.
3. Set resolution to 150 DPI, or 200 DPI if any of the conditions above apply.
4. Tick Convert to greyscale for a black-and-white document.
5. Leave quality at 75%.
6. Verify before you delete anything (next section).

If it is still too large, go back to the original file and try harder settings — not the compressed output. Compressing an already-compressed scan re-encodes degraded images and the artefacts compound visibly for very little extra saving.

Every step here is lossy and irreversible. Downsampling discards pixels permanently. Keep the original until the compressed version has been checked and accepted.

The verification routine

Four checks, two minutes, before anything is submitted or the original deleted.

1. Open at 100% zoom, not fit-to-width. Fit-to-width scales the page down and hides the damage — a badly compressed scan often looks perfect that way. Set the zoom to exactly 100% and read a full line.

2. Go to the page with the smallest text. Not a representative page. Find the footnotes, the terms section, the handwritten box. That page determines whether the file is acceptable.

3. Read a reference number out loud. Account numbers, policy numbers, dates. If you have to look twice at any character, the compression went too far. This is the single most reliable test, because it fails before general readability does.

4. Check the signature. Signatures are thin, variable strokes and they are often what the recipient cares about most. At 96 DPI a signature frequently becomes a smudge.

If any check fails, go back to the original and compress at one setting higher.

Myths worth discarding

• “Greyscale makes it harder to read.” It does not. Readability is set by resolution and JPEG quality. On a black-and-white document, greyscale removes information that was never there.
• “Compress twice for a smaller file.” Each pass re-encodes already-degraded images. Two mild passes look worse than one firm pass and save less.
• “Zoom in to check quality.” Zooming to 400% tests the pixel data, not the reading experience. It makes every compressed scan look bad and tells you nothing useful. Check at 100%.
• “If it looks fine on my phone it is fine.” Phone screens are small and dense and flatter compressed images considerably. Check on the largest screen you have, and assume the recipient may print it.
• “OCR will fix a low-resolution scan.” OCR reads the pixels you give it. Below 200 DPI the recognition error rate climbs sharply, and OCR errors are worse than blurry text because they look confidently correct.
• “Black-and-white mode is always smallest and sharpest.” Only on clean printed text. On anything with shading it is both ugly and sometimes larger, because dithering patterns compress poorly.

What good looks like

A realistic set of numbers for a 12-page contract scanned at 600 DPI in colour:

The useful part of that table is the middle: 150 DPI greyscale is a 97% reduction with no practical readability cost. Everything below it buys a little more space for a disproportionate amount of damage.

If you still have the paper, re-scanning at 200 DPI greyscale beats compressing a 600 DPI colour scan — you avoid creating the excess data rather than discarding it, and the result is visibly cleaner. For phone scanning, scan documents to PDF with your phone covers the settings that matter.