How to split a PDF, and why the parts can weigh more than the whole

Each part is a whole PDF

When Split PDF makes a part, it starts a new PDF and copies into it the pages of that part, with everything those pages use: the fonts, the pictures, the links and the form fields that sit on them. A picture that sits on a single page goes into a single part. A font used on every page goes into every part.

That last point is the whole story. A PDF stores each font once and lets every page share it. Split it into one file per page, and each file needs its own copy.

Where the parts grew: fonts

Our lecture notes have 117 pages set in many embedded fonts. Split into 117 parts, they added up to 31.1 MB, 5.8 times the original 5.3 MB. Almost all of the growth was fonts: the whole file holds 636 KB of font data, and the parts together held 25.3 MB, 39.8 times as much.

The pictures did not grow at all. Each one sits on a single page, so it went into a single part, and the parts held 3.5 MB of pictures, exactly what the whole file holds. Adobe notes that adding a single font to a PDF can sometimes increase its size substantially; splitting can repeat that cost in every part that uses the font.

Scans: what every part repeats

A scanned document is one picture per page, and each picture goes into one part. Our scanned NASA report has 150 pages; split into 150 parts, it added up to 4.8 MB, 32% more than the whole, with the pictures the same bytes as before. Most of the growth, 80%, was the metadata of the document: a description of the file in a format called XMP, 6.2 KB in the original. Split PDF copies it into every part, so the parts held 930 KB of it. The rest is the small structure each part needs around its page.

Our PDF of photos carries no such description. Split into its 4 pages, it grew by only 1,304 bytes.

When the parts weigh less

The IRS form went the other way. Split into its 2 pages, the parts added up to 98.0 KB, 33% less than the original. The form carries a second description of itself, in a format called XFA, that only some programs read. Split PDF does not copy it into the parts, and it says so next to the file; the form fields themselves are kept.

The parts are also written with their structure packed tightly, in compressed object streams, which is part of why a part can come out smaller than its share of the whole.

How to split without the extra weight

Split into ranges, not single pages, when the pages share their fonts. A chapter in one part carries its fonts once, where its pages split apart carry them once each. Type the ranges you need in the tool, for example by chapter or by section, instead of splitting every page.

If you split to get under an email limit, check the total, not only each part: a part that fits can still be the first of many heavy ones. Each part of our scanned report added about 7.7 KB to the total, and 6.2 KB of that is its own copy of the metadata; the rest is its structure. Fewer, larger parts repeat both fewer times. Compress PDF can make the pictures lighter before you split.

And if what you need is a shorter PDF rather than several, delete the pages you do not need instead. One file keeps its fonts once, however many pages go.

What we measured on each file

We split 5 PDFs into one part per page, the default of Split PDF, and added up the parts. The parts weighed more than the whole in 3 of them. The table shows each file, with its fonts before and after, since fonts made most of the extra weight in our sample.

What we measured

Each PDF split into one part per page, the default of Split PDF. Change is all the parts against the whole file; fonts shows the font data of the whole file, then of all the parts together.

Measured on November 2, 2026 in Node with the engine of this site, which gives the same bytes as the browser, on five PDFs of our tests, the real documents and a PDF of photos. Fonts, pictures and metadata were counted with pdf-lib, the library the engine opens files with.

Sources

• Adobe: Why is my PDF so big?. Read on November 1, 2026.