Rotate the pages of a PDF

Two different things are both called «a crooked page»

A PDF reader shows them to you identically, but they come about in two opposite ways, and it is worth knowing because it explains why certain tools «do not work». In the first case the page carries inside itself how far it should be turned: the content is drawn straight and there is a separate instruction telling the reader to show it rotated. That is what a scanner does when it works out the orientation by itself.

In the second case the page declares itself straight and is not: the content is actually drawn sideways, and there is no instruction in the file to correct. That is the badly made scan, and it is the case people really have in their hands. This page sorts out both, because to work out the orientation it does not look at the instruction: it looks at which way the writing runs.

How it works out the orientation on its own, and when it says nothing

Text inside a PDF is not a picture of letters: every piece carries with it the direction in which it runs. By adding up how many characters run in each of the four directions you know where the top is, with no optical recognition and without downloading anything. On test sheets built on purpose, with the orientation known from the start, the count got 33 pages right out of 33 and got none of them wrong.

But there are pages about which nothing can be said, and then nothing is said: a blank page has no orientation, one with two words does not have enough of one, and a table with turned headings has writing in several directions. Above all, a scan with no text inside is only a picture: there is no writing there to read, the page tells you so and you turn them by looking at them. A recognition that guessed in those cases too would turn landscape tables by mistake, and those are landscape on purpose.

Turning a page does not touch what is inside it

The rotation does not redraw and does not recompress anything: the instruction saying how to show that page is added to the document. The text stays text, so it can still be selected and searched; the images stay as they were, with the same quality and the same weight; and the file that comes out is the document from before, turned. That is the difference between this and turning the pages into photographs, which instead takes the search away and bloats the file.

For the same reason a page that is already straight does not get touched: if you gave it no turn, nothing gets written for it. And if you change your mind, the «Put back as it was» button resets everything without your having to load the file again.

The tools next door

If as well as turning them you also want to reorder them or delete some, the right place is Organise a PDF. If the PDF is a scan and you would like to be able to search for words inside it, the step to take first is Make a scanned PDF searchable, which puts the text underneath the images.

If instead the photo of a document is crooked because you took it at an angle, that is not a rotation problem but a perspective one, and it gets straightened with Straighten a photographed document. And to pull the pages out as image files there is PDF to images.