Split PDF

Inside every PDF there is an object called the page tree. It lists each page and points at the content stream that draws it — the text operators, the embedded fonts, the image XObjects. Reordering, deleting or extracting pages rewrites that catalogue; it does not redraw anything.

That distinction matters in practice. Because the drawing instructions are copied untouched, an organised document keeps exactly the visual fidelity of the original: vector text stays vector, 600 DPI scans stay 600 DPI, and no generation loss creeps in no matter how many times you split and recombine a file.

People are often surprised that deleting half the pages barely shrinks the file. The reason is shared resources: fonts, colour profiles and repeated logo images live in a resource dictionary that many pages reference. Removing the pages does not automatically remove the objects they pointed at.

If size is your real goal, organise first and then run the result through a compression pass, which rewrites image streams and drops orphaned objects. Doing it in that order avoids compressing pages you were about to throw away.

Two properties travel with each page and cause most of the confusion in reorganised documents. The first is the rotation flag, an integer of 0, 90, 180 or 270 that viewers apply at display time — a page can look upright while its underlying content is sideways. The second is the MediaBox, the page's physical size.

A document assembled from several sources can legitimately contain A4, Letter and a wide landscape spreadsheet page. Nothing is broken; viewers simply render each page at its own dimensions. Normalise them deliberately with a resize step if a print shop asks for a single trim size.