Why is my PDF so big?

Two PDFs with the same number of pages can differ in size by a factor of a thousand. Our largest test file weighed 5.9 MB per page and our smallest less than 5 KB per page, and both were ordinary documents. The difference is always in what the file carries, and you can usually tell which kind of file you have in a few seconds.

What is inside a PDF

A PDF is a container. For each page it stores a list of drawing instructions (put this text here in this font, draw a line there, place this picture) and the things those instructions use: fonts and pictures. Everything else, such as the page list, bookmarks and the index at the end of the file, is usually small.

We wrote a short script that goes through every object in a PDF and adds up the bytes of each kind, then ran it on five files.

Cause 1: photos at camera resolution

The brochure is 8 pages of text with four photos, each pasted in straight from a camera. That is what happens when you drag a photo into most document editors: the program keeps every pixel, even when the photo is shown the size of a postcard.

Each photo is shown 6.6 inches wide. At 150 pixels per inch, plenty for a screen or an office printer, that is about 1,000 pixels across. The photos had about 6,000, so they carried 36 times more pixels than the page can show. Resampled to 150 dpi with our PDF compressor, the four photos took less than 280 KB together, and the whole brochure went from 47.3 MB to 287 KB.

Cause 2: scans

A scanned document has no text at all, only one picture per page. Our scan of a 10 page report, at 300 dpi in colour, came to 9.4 MB, or 961 KB per page. The same report saved as a normal PDF was 47 KB, about 200 times smaller, because text is stored as characters rather than as millions of pixels.

Most scanners and scanning apps default to 300 dpi, which is right for archives and print but more than a screen needs. Reduced to 150 dpi, the scan was 2.2 MB and the text read just as well. There are close-ups of each setting in our compression test. If you scan documents often, setting the scanner to 150 or 200 dpi saves the step.

A scan also cannot be searched or copied from. If you need the words, OCR reads them and adds an invisible text layer, so the pages look the same but the text can be selected.

Cause 3: fonts embedded in full

The report is 11 pages of plain text with headings and no pictures, and it is still 745 KB. 89 percent of that is two fonts, Georgia for the text and Arial Bold for the headings, embedded in full. A font file holds every character it supports, often thousands of them for dozens of languages. Arial Bold on Windows 11 is a 967 KB file on its own, and here it draws a handful of headings.

Most programs avoid this by embedding a subset: only the characters the document uses. A subset is usually a few dozen kilobytes. When a PDF that is only text is unexpectedly large, the fonts were probably embedded whole.

A compressor cannot fix this without changing how the text looks, which is why ours leaves fonts alone. The fix is at the source: export the document again with font subsetting on (the default in Word, LibreOffice and the print to PDF of web browsers), or choose a standard font.

Cause 4: drawings and hidden extras

Our fifth file is a 14 page research paper with charts and diagrams, the one Mozilla uses as the demo for pdf.js. It is the most balanced of the five: 49 percent page content, mostly the drawing instructions for its charts, 10 percent pictures, 9 percent fonts and the rest structure. At 993 KB it is a well made PDF, and there is little any tool could remove.

Vector drawings are usually small, but there are exceptions. Maps, building plans and charts with hundreds of thousands of points can make page content the biggest part of a file. If a file like that is too big, a simpler export from the program that made it usually works better than a compressor.

A few other things add weight without showing on any page:

• Edits saved on top of the file. PDF allows a program to append changes to the end of a file instead of rewriting it, so every earlier version stays inside. Saving a fresh copy, or running the file through a tool that rewrites it, drops them.
• Attachments. A PDF can carry other files, such as spreadsheets or source documents, listed in the attachments panel of the viewer.
• The same picture stored many times. A logo on every page should be stored once, but some programs store a copy for each page.

How to tell what is in your PDF

You do not need a script. Three quick checks narrow it down:

1. Divide the size by the number of pages. A few dozen KB per page is normal for text. Several hundred KB or more per page means pictures or scans.
2. Try to select a sentence. If you cannot, every page is a picture: a scan or a flattened file.
3. Zoom in to 400 percent on a photo. If it stays sharp far beyond what the page needs, the photos are at full resolution, and compressing them will help a lot.

If the size per page is small and the text selects normally, the file is probably fine as it is. If it is still too big for an upload limit, splitting it into parts is often the only real option.

What to do about each cause