A scanned PDF is not a document in the way a Word export is. It is a stack of photographs with a PDF wrapper around them. Every page is one big image, and the file size is simply the sum of those images. That single fact explains almost everything about why scans are huge and why some compression tools barely change them.
A colour A4 page scanned at 300 DPI is about 8.7 megapixels. Stored as a high-quality JPEG, that is commonly 3 to 8 MB. Six pages and you are at 30 MB before anything else is added. The text on the page contributes nothing to the size, because there is no text — only pixels that look like text.
Why normal compression barely helps
Most PDF compressors do one of two things. The gentle kind rewrites the file structure, deduplicates fonts and packs objects more tightly. On a PDF exported from Word that can save 30–50%, because there is a lot of structure to clean up. On a scan it saves 1–3%, because the images were already compressed by the scanner and there is nothing else in the file.
The stronger kind re-renders each page at a lower resolution and saves it again as a JPEG. That is where real savings come from — and it is also where documents get ruined, because the setting that makes the file small is the same setting that turns 9-point text into grey mush.
What actually makes a scan smaller
Only two levers matter: how many pixels each page has, and how aggressively those pixels are compressed.
| Resolution | What it is good for | Rough size per A4 page |
|---|---|---|
| 300 DPI | Printing, archiving, fine print and stamps | 3–8 MB |
| 200 DPI | OCR, reading on a laptop, most official uploads | 0.8–2 MB |
| 150 DPI | Reading on screen, email, sharing with a friend | 0.4–1 MB |
| 96–120 DPI | Thumbnails, previews, quick reference copies | 0.1–0.4 MB |
In the Compress PDF tool the three modes map to this directly. Light is the structural pass: it keeps text as text and is the safest option, but on a scan expect very little change. Balanced re-renders pages at 140 DPI with moderate JPEG quality, which is the usual sweet spot for a scan you need to read. Strong drops to 96 DPI and lower quality — small files, visibly softer text, fine for a preview and wrong for anything official.
The trap: OCR text layers
If a scanned PDF has been through OCR, it contains two things per page: the image, and an invisible layer of real text sitting on top of it. That text layer is what makes the document searchable, selectable and readable by screen readers. It is also what most government and university portals use to check that you uploaded a real document.
Re-rendering a page throws that layer away. The result looks identical and is no longer searchable. If the text layer matters, either use Light, or compress first and run OCR on the compressed copy afterwards. Do not compress an OCR'd document with Strong and then wonder why the text cannot be selected.
A workflow that does not ruin the document
- Clean up first. Blank backs of pages, duplicate scans and cover sheets are pure weight. Remove them with Delete PDF pages before compressing — deleting three blank pages from a twelve-page scan is a 25% saving for free.
- Compress a copy, never the original. Keep the untouched scan. Every re-render is lossy, and doing it twice compounds the damage.
- Start with Balanced. Check the page count, open the result and read a dense paragraph at 100% zoom, not in a thumbnail. If the text is still crisp, you are done.
- Only then try Strong. Compare the two files side by side. If Strong makes a signature or a stamp illegible, it is the wrong choice regardless of the file size.
- Still too big? Split the document into parts with Split PDF, or scan again in grayscale. Converting a colour scan of black text to grayscale typically cuts the image data by two thirds, and it is better done at the scanner than after the fact.
Things that do not work
- Zipping it. A PDF full of JPEGs is already compressed. A ZIP will save a fraction of a percent, and most upload forms do not accept ZIPs anyway.
- Printing to PDF again. This often makes scans larger, because the printer driver re-encodes the images at its own settings.
- Screenshotting each page. It works, in the sense that the file gets smaller. It also destroys the resolution, the page size and any text layer, and turns a 6-page document into an unprofessional mess.
- Renaming the file. The size is inside the file. The name has nothing to do with it.
How small should it be?
Work backwards from the destination. Most email services cap attachments at 25 MB. University and government portals commonly cap at 2 MB, 5 MB or 10 MB, and a few are stricter. Printing shops and archives want 300 DPI and do not care about size. Decide which of those you are doing, pick the resolution from the table above, and stop there — there is no prize for the smallest possible file if the reader cannot read it.
One more reason to do it locally
Scanned documents tend to be the sensitive ones: ID cards, bank statements, medical records, signed contracts. Uploading them to a free compression site means handing a copy to a stranger's server, where it may sit in a queue, a cache or a backup you cannot see. Every tool linked here runs in the browser tab, so the file is processed on your own device and never uploaded. That is not a marketing line — it is the reason the tools exist.