Compress a scanned PDF without wrecking the text

Steps, on a real 10-page scan

The numbers below are from our test file: an invented 10-page lease scanned at 300 dpi in colour, 17.6 MB, about 1.8 MB a page.

1. If the scan also needs to be searchable, do that first. Run OCR on the sharp original and compress the OCR result. OCR reads 300 dpi far better than 110 dpi, and the text layer it adds is text, which compression keeps exactly as it is.
2. Drop the scan into the compressor above (or open /compress-pdf). Several files at once are fine; each is compressed on its own.
3. Pick a preset. Email is selected by default: pictures at 110 dpi, JPEG quality 65. Switch to Web (150 dpi, quality 75) if the reader may zoom in on details.
4. Or type a size target when a portal or form names a number, in KB or MB. The tool tries its settings from 300 dpi down to 50 dpi and keeps the gentlest one that fits, whatever the preset says. A file already under the target comes back untouched.
5. Press Compress and read the result. It shows the size before and after, how many pictures were re-encoded and to what dpi, and the 5 pages that changed most, each with its picture bytes before and after.
6. Open the result and zoom to 200 percent on the smallest text, any stamp and any signature. If they still read, send it. If not, go back one step: Web instead of Email, or a looser target.

What each preset did to the 10-page scan:

Preset	Pictures	Result	Saved
Print	300 dpi, quality 85	17.6 MB, returned unchanged	0, and the tool says why
Web	150 dpi, quality 75	3.2 MB	82 %
Email	110 dpi, quality 65	1.7 MB	90 %

The Email preset took about 3 seconds for all 10 pages on a mid-range 2024 laptop.

Pick the resolution by what the file is for

Resolution is the setting that matters most on a scan, because the pixel count falls with the square of the dpi. Reduce PDF DPI explains the math; here is the short version for scans:

• 110 dpi (Email): email, portal uploads, anything read once on a screen at page width. Normal-sized print reads cleanly.
• 150 dpi (Web): files people will zoom into, forms with small boxes, contracts someone will read closely on screen.
• 300 dpi (Print): anything you may print, archive or need to zoom into later. Keep that one as your original rather than compressing it.

If you are not sure, start with Email, look at the result, and step up only if the smallest text breaks up.

Grayscale: try it, but do not expect miracles

Most scanned paperwork is black ink on white paper, scanned in colour. It sounds like dropping the colour should be the big saving. On our tests it is not:

10-page scan	Size
Original, colour	17.6 MB
Grayscale PDF only	15.4 MB
Email preset only	1.7 MB
Grayscale, then Email preset	1.7 MB

JPEG already spends few bytes on the colour of plain paper, so once the resolution drops, grayscale adds little. It helps more on colour photos you do not need in colour, and it makes a copy print predictably on a black-and-white printer.

Keep the colour when it carries meaning: a company stamp, a signature in blue ink, a highlighted clause, a chart keyed by colour. Those do not survive grayscale.

Hitting a portal’s size cap with a multi-page scan

Portals and application forms often name an exact number. Type it as the size target, or open the matching page directly: compress to 1 MB, 500 KB, 200 KB, 100 KB or 50 KB. The targets count 1 KB as 1,000 bytes, the stricter reading, so the file passes whichever way the portal counts.

The same 10-page scan against each target:

Target	Result	Setting it ended on
2 MB	1.8 MB	110 dpi
1 MB	637 KB	72 dpi
500 KB	366 KB	60 dpi
200 KB	197 KB	50 dpi, quality 25
100 KB	missed: 198 KB	the strongest setting, and the tool says so

A rule of thumb from these: a scanned A4 page costs about 180 KB at 110 dpi, about 60 KB at 72 dpi and about 19 KB at the strongest setting. So a 100 KB cap fits about 5 scanned pages and a 200 KB cap about 10, and the 5 pages will look far better than 10 would.

When the cap is that tight, send fewer pages: extract the pages the receiver needs, or split the scan into parts if the portal takes several uploads. Cut to 5 pages, our scan reaches 96 KB on the 100 KB target.

What compression cannot do for a scan

It is lossy and one-way. Re-encoding changes pixels, and detail that is dropped cannot be restored by any tool. The compressor writes a new file (lease-compressed.pdf) and leaves yours alone; keep the original scan, send the small copy.

It does not make the scan searchable. A compressed scan is still pictures of pages. That is what OCR is for, and it belongs before compression, not after. The scanned PDF to searchable guide covers it.

It leaves 1-bit scans alone. A black-and-white scan stored at 1 bit per pixel (fax style) is already compact, and the compressor does not touch it.

It removes a password. A protected scan opens when you type its password, and the result has none; the tool warns you. Protect the copy again with Protect PDF if it needs one.

Some caps cannot be met legibly. 10 colour pages at 100 KB is about 10 KB a page, and no setting keeps text readable at that size. When the tool misses a target it says so, with the size its strongest setting reached and the reason (“Most of what is left is the page pictures at the lowest quality this tool offers; fewer pages gets further”). The honest fixes are fewer pages or a scan made at 150 dpi in grayscale in the first place. Why scanned PDFs are huge explains the math behind that limit.