Split PDF pages in half

The problem this exists for

You scan a book, or download one, and every page of the PDF holds two pages of the book side by side. It is unreadable on a phone, because the reader fits the whole spread to the screen width and each page ends up half the size it should be. It prints two-to-a-sheet whether you want that or not. And the page numbers in the PDF no longer match the page numbers in the book.

Cutting each spread into two pages fixes all three at once, and it is the single most useful thing you can do to a scanned book before reading it.

The cut is worked out from the page shape

A spread is wider than it is tall, so a landscape page is two portrait pages side by side and gets cut down the middle. A portrait page that holds two is the stacked case and gets cut across. Auto picks the right one from the proportions, which is correct for a scanned book almost every time — but you can force either.

Reading order, for the scripts that need it

A book in Arabic, Hebrew, Urdu, Persian or Japanese manga opens the other way, so the right-hand half of a spread is the page that reads first. Getting that wrong means every pair in the document is reversed, and fixing it afterwards means reordering hundreds of pages by hand.

So it is a setting, and it is worth checking before you convert a long book rather than after.

Overlap, for what sits in the gutter

A clean cut down the exact middle slices through anything that crosses the fold — a wide table, a diagram spanning both pages, a line of text on a badly-bound book that runs into the crease.

Keeping a strip of the gutter on both halves means whatever is there appears on both pages rather than being cut in two. Twenty to thirty points, which is about a third of a centimetre, is usually enough. It is not free: you get a duplicated sliver down one edge of each page.

Moving the cut, for a scan that was not square

Books do not sit flat on a scanner. If the spine was off-centre, the exact middle of the image is not the middle of the book, and a centred cut takes a strip of the left page onto the right one all the way through the document.

The offset shifts the cut away from centre — positive moves it right, or down for a horizontal cut. Open one page at 100%, read off how far the fold actually is from the middle, and use that. It is measured in points, 72 to the inch.

Cropping first is often the better order

If the scan has a wide white surround as well as two pages, crop the margins first and then split. The cut here is measured from the boundary the page actually displays at, not from an invisible larger sheet, so it lands where you would expect on an already-cropped page — verified, on a page whose boundary was inset by 100 points, the cut landed at the displayed centre rather than the sheet's.

Nothing is re-rendered

Both halves point at the same content that was already in the file; only the visible boundary differs. So there is no quality loss at all — no re-compression, no re-rasterising, and barely any increase in file size, because the page content is stored once and referenced twice.

The other consequence is that the halves are not really separate images: the content of the whole spread is still there on each one, hidden outside the boundary. That is fine for reading and printing. It is not a way to remove something from one half.

Leave some pages alone

A cover, a title page, a full-page map or a photographic plate is one page, not two, and cutting it in half ruins it. Name the pages you want cut and the rest pass through untouched.

Both the page boundary and the physical page size are moved, which matters for printing: a viewer shows the boundary, but a printer reads the page size, and leaving that at the full spread makes each half print at spread size with the other half as blank paper.