Why a compressed PDF is still too large
Checked against our tools on 19 October 2026
PDF compression works by re-encoding images. If a PDF is mostly text, its images are already compressed, or Light and Medium are leaving its screenshots alone to protect quality, there's little to remove. And when it did shrink but is still over a limit, the usual cause is page count: even at Strong, each scanned page in our tests still weighed about 90 KB. Try Strong, then remove pages nobody needs or split the file. Our compressor never hands you a bigger file: if the result isn't smaller, you keep the original.
Go straight to the tool: Compress PDF, Remove pages, Split PDF
What compression can and can't change
Almost all of the weight in a large PDF is images: scanned pages, photos, screenshots. Text, by comparison, is tiny. A page of text is a few kilobytes because the PDF stores the characters and a reference to a font, not a picture of the letters.
So a compressor has two jobs. It re-encodes images at a lower quality or resolution, which is where the big savings are, and it tidies the file's internal structure, which saves a little on any file. It doesn't touch text, fonts or vector drawings, so those never get blurry.
Tested: four kinds of PDF
We ran our compressor on four sample files at each of its three levels:
| Sample | Original | Light | Medium | Strong |
|---|---|---|---|---|
| 12-page text document | 26.0 KB | 13.6 KB | 13.6 KB | 13.6 KB |
| 3-page office scan (300 DPI JPEG) | 2.42 MB | No real change | 533 KB | 266 KB |
| 1-page report with a PNG screenshot | 2.28 MB | Not reduced | Not reduced | 90 KB |
| 4 pages of large photos | 12.49 MB | Not reduced | 3.22 MB | 272 KB |
Reading the results
The text document shrank by the same amount at every level, because the only saving was structural. There were no images to re-encode, so Medium and Strong had nothing extra to do.
The scan and the photos didn't change at Light. Light is lossless: it never re-encodes an image, so it can only tidy structure, and these files were already almost all JPEG data. At Medium and Strong the images were re-encoded and shrank by 74% to 98%.
The screenshot is the interesting one. It's stored losslessly, like a PNG, which is the right choice for sharp text and interface edges. Light and Medium leave lossless images alone because turning them into JPEG can smear fine detail. Strong converts them anyway, which is why the file went from 2.28 MB to 90 KB, with softer edges.
Other reasons a PDF won't shrink
Our compressor also leaves some images exactly as they are, because re-encoding them could change how they look:
- Images with transparency or masks.
- Images in CMYK or other print colour spaces, rather than RGB or greyscale.
- Images with a bit depth other than 8 bits per channel, including 1-bit black-and-white scans, which are already very compact.
- Images that would shrink by less than 10%, where the quality cost isn't worth it.
It got smaller, but not small enough
Compression has a floor, and it's set by the number of pages. Our 3-page scan came out at 266 KB on Strong, about 90 KB a page. A 30-page scan compressed the same way lands near 2.7 MB, so it won't fit a 2 MB upload form at any level without losing pages. Photos behave the same way: Strong brings each one down to 1,500 pixels on the long side, and below that, text in a photographed document stops being readable.
The limit is often tighter than it looks, too. Email encodes attachments as text on the way, which adds about a third, so a 19 MB PDF can bounce off a 20 MB limit. Sizes aren't even counted the same way everywhere: we and Windows count a megabyte as 1,048,576 bytes, while a Mac's Finder counts 1,000,000, so a file we call 2.00 MB shows as 2.1 MB in Finder. If you're within a few percent of a limit, aim lower.
If the file is mostly text and still several megabytes, compression isn't the tool. Something else is taking the space, usually hundreds of pages or large embedded fonts, and removing pages will do more.
What to try instead
- If you have a limit to hit, choose Get it under a size in Compress PDF and enter the limit. It tries the gentlest settings first and stops at the first one that fits, and if nothing fits it says so and gives you the smallest readable version. Our guide on getting a PDF under an upload limit covers it.
- Try Strong, and check the result at the zoom you'll read it at. Text and fonts are never re-encoded, so typed text stays sharp; it's scanned and photographed text that softens.
- Remove pages nobody needs: cover sheets, blank backs of scanned pages, appendices the recipient already has. Each page you drop takes its images with it, and Remove pages runs in your browser.
- Split the file into parts that each fit the limit with Split PDF, and name them so the order is obvious, such as report-part-1-of-2.pdf.
- If you made the scan yourself, scan again at 200 DPI in greyscale. That usually gives a much smaller file than compressing a 600 DPI colour scan afterwards, and it's still plenty for a signed form or a letter.