Cut scanned spreads into single pages

How to Split PDF in Half

1. Add the scanned PDF. Drop in the file of spreads. It is read on your device and never uploaded.
2. Choose the direction of the cut. Down the middle for pages photographed side by side, across the middle for pages stacked one above the other.
3. Move the cut to the fold. Nudge the slider until the cut sits on the gutter. A book held open under a camera rarely lands at exactly fifty percent.
4. Leave the covers out. Deselect any single pages — front and back covers usually — and they pass through whole.

Why the fold is never at fifty percent

This is the detail that separates a usable result from forty pages of ruined text. A book photographed open does not present two equal halves. The spine sits wherever the book falls open, which shifts with the binding, the thickness of the block and how far through it you are, and the camera is rarely square to the page.

A fixed 50% cut on a scan whose fold sits at 47% slices a narrow strip off the inner margin of every left page and glues it to the right one. On a wide margin nobody notices. On a tight one it takes a column of text with it, on every page.

So the position is a control rather than an assumption. Set it once against a representative page and it applies throughout — a scan produced in one sitting is consistent even when it is not centred.

Rotation is where most implementations go wrong

A spread photographed on a phone is very often stored as a portrait page carrying a rotation flag, rather than as a genuinely landscape page. The viewer honours the flag, so it looks landscape on screen; the file underneath is portrait.

A tool that cuts along the page's stored horizontal axis will therefore cut that spread into a top and a bottom half, not a left and a right one — on exactly the input the tool exists for, and with no error to explain it.

The cut here is worked out in display space, meaning it always falls where you can see it falls, and is then converted back into the page's stored coordinates. Rotated, unrotated and upside-down pages all behave the same way.

Right-to-left books

Japanese manga, and books set in Arabic or Hebrew, are read from right to left, so on a scanned spread the right-hand page comes first. Cutting a spread with the usual assumption produces a correctly-cut document in exactly the wrong order.

Turning on right-to-left reading order swaps which half is emitted first, so the resulting pages run in the order a reader of that language expects.

What the cut does to quality

Each half is a crop of the original page rather than a re-render. The underlying content is untouched and no image is recompressed, so a scan keeps exactly the pixels it arrived with and a text PDF keeps selectable text.

That matters for archival scans in particular. Anything that decodes and re-encodes a JPEG loses a little each time, and a workflow of scan, split, deskew, OCR would compound that loss at every step if each stage rasterised. None of them do.

Bookmarks are remapped to the first half of the page they used to point at, so an outline added before splitting still lands in the right place afterwards.

Frequently asked questions