You scan three pages and end up with a 40 MB file that your email client refuses to send. Meanwhile a hundred-page technical manual downloads in a second and takes up 2 MB.
The difference isn’t the page count. It’s what’s on the pages — and once you know that, both the problem and the fix become obvious.
Two completely different kinds of PDF
Almost every oversized PDF is oversized for the same reason, so let’s start there.
A text PDF — one made by exporting from Word, or by a report generator — stores letters as letters. “The” is stored as three characters plus instructions about which font and where to place them. This is astonishingly compact. It’s why a long manual can be smaller than a single photograph.
A scanned PDF is a photograph of a page. Every letter is just dark pixels among light ones, and the file has no idea any of it is text. A page scanned at high resolution in full colour can easily be 5–15 MB on its own.
Test which one you have: open the file and try to select a sentence with your mouse. If you can highlight words, it’s text. If you can only draw a box across the page, it’s an image — and that’s where the megabytes went.
The four things that make files enormous
Scanning in colour when you didn’t need to. A black-and-white page scanned in full colour stores three colour values per pixel to represent what’s essentially ink or no ink. This is the single biggest cause of huge files, and it’s usually a default setting nobody changed.
Resolution far beyond what’s useful. 600 dpi is for archival reproduction. For a document someone will read on screen or print on an office printer, 200–300 dpi is indistinguishable and a fraction of the size.
Images inserted at full camera resolution. Drop a 12-megapixel photo into a report and it usually gets embedded at full size, even though it’s displayed at 8 cm wide. The file carries all those pixels regardless.
Embedded fonts. Fonts are stored inside the PDF so it looks right everywhere. Normally that costs a few hundred kilobytes — but a document using many fonts and embedding them completely rather than only the characters used can add several megabytes.
What compression actually does
Two mechanisms, and the difference matters.
Lossless compression packs the data more efficiently without discarding anything. Unpack it and you get back exactly what went in. This is what happens to text and structure, and there’s no reason not to use it.
Lossy compression throws information away permanently, choosing details your eye is unlikely to miss. This is what happens to images, and it’s where the real savings are. It’s also irreversible — no tool recovers what was discarded, so keep your original.
Push lossy compression too hard on a scanned document and the results are ugly in a specific way: text edges go fuzzy, faint stamps and handwriting disappear, and thin lines in tables break up. On a scanned signature or a form filled in with a pen, over-compression can destroy the very thing that made the document worth keeping.
How to actually shrink a file
In the order that gives the most benefit for the least damage:
1. Fix it at the scanner. Easily the best option. Scan in greyscale or black-and-white unless colour carries meaning, at 200–300 dpi. A 40 MB scan often becomes 2 MB, with no loss of anything you’d notice.
2. Remove pages you don’t need. Obvious and frequently forgotten. Sending 4 relevant pages instead of a 60-page report is a bigger saving than any compression setting, and a kindness to the recipient.
3. Downsample the images. Most compression tools let you cap image resolution. Setting it to 150 dpi for screen reading or 300 dpi for printing typically removes most of the bulk.
4. Then compress. Start moderate. Check the result before accepting it — particularly any small print, stamps or handwriting.
And always check the output. Compression is the one operation where the damage is invisible until someone needs to read a figure that’s now a grey smudge.
The trick that beats compression entirely
For scanned text documents there’s an option most people never consider, and it often works better than any compression setting: run OCR and keep the text.
OCR software reads the image and works out what the letters are. Some tools then produce a “searchable PDF” that keeps the scan and adds an invisible text layer — useful, but no smaller. Others can rebuild the document as real text, and that’s where the size collapses. A 30 MB scan can become a few hundred kilobytes, because you’ve stopped storing a photograph of words and started storing the words.
The trade-off is honest and worth stating: you lose the visual fidelity of the original, including handwriting, stamps and the reassuring look of a real document. OCR also makes mistakes, so the text needs checking. That makes this a poor choice for anything where the appearance is part of the evidence — and an excellent one for a scanned manual, a set of notes, or a report you only need to read and search.
A middle path many people miss: keep the scan for the record, produce a text version for circulation. The 30 MB original goes in the archive; the small readable one goes in the email.
When not to compress
Some documents shouldn’t be squeezed at all.
Anything legally significant. Contracts, notarised documents, evidence. If a signature or a stamp has to be legible and verifiable, don’t put it through lossy compression.
Signed PDFs. Compression rewrites the file, which invalidates a digital signature. Not a bug — the signature working correctly. But if the signature matters, the file has to stay exactly as it is.
Archival copies. Compress the version you send. Keep the original for the archive. Storage is cheap; a detail you can’t recover isn’t.
A note on where you compress
Compression is one of the operations people most reliably do online, because it’s the one where you’re already frustrated — the email bounced, you want it fixed now.
That frustration is worth noticing, because it’s exactly when people stop thinking about what they’re uploading. The files that need compressing are typically scans, and scans are typically documents that came on paper: signed contracts, identity documents, medical results, bank statements. Those are the most sensitive documents most people handle, and “the email bounced” is a poor reason to put one on a stranger’s server.
PDF Manipulator compresses locally, so the scan of your passport or your signed lease never leaves your machine. It won’t perform magic — the physics of lossy compression are the same everywhere, and a document scanned badly at 600 dpi in colour will always be a compromise. But the compromise stays between you and your own computer.
The short version
PDFs are big because of pictures, not pages. Check whether yours is text or a scan, fix the scanner settings if you can, drop the pages nobody needs, and compress moderately rather than aggressively. Keep the original, never compress a signed document, and give a moment’s thought to what’s on the page before uploading it somewhere to save eight megabytes.
Compress PDFs on your own computer — PDF Manipulator is free →
Sources
- Artifex, Ghostscript: High Level Devices (PDFSETTINGS presets)
- Tesseract OCR
- ISO, ISO 32000-2:2020 — Document management, Portable Document Format, Part 2: PDF 2.0
