Why a PDF is 40 MB, and how to shrink it without wrecking it
A ten page document should not be forty megabytes. What is taking up the space is almost always the same thing, and the fix depends on what the file is for.
By Hasnain Butt Akhtar, founder and developer ·
If a PDF is unexpectedly large, there is a good chance it contains no text at all. That single fact explains most oversized documents, and it changes which fix will work.
Two completely different things called PDF
A text PDF stores your words as words. The letters are described by the fonts embedded in the file, so you can select them, copy them, and search for them. A twenty page report of this kind is typically well under a megabyte.
A scanned PDF is a stack of photographs. Your scanner or phone took a picture of each page, and the PDF is just a wrapper around those pictures. You cannot select the text because there is no text, only an image that happens to look like text. A twenty page scan can easily be thirty or forty megabytes.
Telling them apart takes two seconds. Open the file and try to select a word. If a text cursor appears and you can highlight it, it is a text PDF. If your pointer draws a box over the page instead, it is a scan.
What is taking the space
| Cause | How to spot it | Typical saving |
|---|---|---|
| Scanned pages at high resolution | Cannot select text, size grows evenly with page count | Very large, often 80 per cent or more |
| Photographs placed at full camera resolution | A few pages are enormous, the rest are small | Large |
| Screenshots saved as lossless images | Documentation and manuals, size out of proportion to content | Moderate to large |
| Every font fully embedded | A text-only file that is still a few megabytes | Small, and risky to change |
| Tracked changes, comments and old revisions | File shrinks a lot when re-saved | Sometimes surprising |
Choose the target before you compress
This is the decision that determines everything else, because compression that is right for one purpose ruins the file for another.
For reading on a screen or emailing, images inside the document only need to be about 150 dots per inch. Anything beyond that is detail nobody will ever see. This is where the dramatic savings come from.
For printing, you need around 300 dots per inch, and compressing below that produces visibly soft, muddy pages. If the document is going to a printer, or to a client who might print it, do not compress it for screen.
For an archive or a legal record, do not compress at all. Keep the original. Store the smaller version separately if you need one for sending.
Most compression tools offer these as presets with names like screen, ebook, print and prepress. Now you know what the names mean, choosing between them is straightforward.
Shrinking a scanned document
Scans have specific tricks available because of what they contain.
- Reduce the resolution to match the purpose. Many scanners default to 600 dots per inch, which is far more than a document page needs. Two hundred is plenty for text, and 300 is generous. This alone often cuts the file by three quarters.
- Scan in greyscale, or black and white, when the original is. A colour scan of a black and white page stores three times the data for no benefit. Most scanners have a document or text mode that handles this properly.
- Let it be slightly lossy. Pages of text tolerate image compression well, because the contrast between ink and paper is so high that small artefacts are invisible. This is not true of photographs or fine artwork, so judge by what is on the page.
- Add text recognition, but expect no saving. Optical character recognition adds a hidden layer of real text over the images, which makes the document searchable and selectable. It is genuinely worth doing. It does not make the file smaller, and it usually adds a little.
Shrinking a document you made yourself
If you produced the PDF from a word processor or design program, the best fix happens before the PDF exists.
Resize images before placing them. A photograph from a modern camera is several thousand pixels wide, and placed into a document at postcard size it is still storing every one of those pixels. Reducing images to roughly the size they will appear, at the resolution you settled on above, prevents the problem rather than treating it.
Then export using the correct preset. Most programs offer a smaller file or minimum size option in their PDF export settings, which does the downsampling for you.
One thing to be careful about: some tools reduce size by flattening the document into images. That makes the text unselectable and unsearchable, turning your clean text PDF into a scan. If you can no longer select words after compressing, that is what happened, and it is worth going back and doing it differently.
What not to do
Do not remove embedded fonts to save space. The saving is small and the risk is real. A PDF without its fonts renders using whatever the reader has available, so line breaks move, tables shift, and a carefully laid out document arrives looking wrong on someone else's machine.
Do not compress the same file repeatedly. The images inside are being re-compressed each time, and the quality drops on every pass exactly as it would with a photograph. Always go back to the original.
Do not compress and then delete the original, particularly for scans of documents you cannot re-scan. Keep the full quality version somewhere, even if you never send it.
A sensible rule of thumb
For a text document with a few images, expect somewhere under a megabyte for ten pages. For a scanned document at sensible settings, expect roughly 100 to 300 kilobytes per page. If you are far above those numbers, something in the file is stored at a much higher quality than the job requires, and the table above will tell you which.
If you are hitting an email limit, remember that most mail systems cut off around 25 megabytes including the encoding overhead, so a 24 megabyte attachment can still bounce. Below 10 megabytes is the comfortable range, and anything larger is usually better sent as a link than squeezed until it looks bad.