Rotate a PDF so that it stays rotated
Most people who land here have already rotated the file once. They turned it in Chrome, or in Preview, or in the free Adobe Reader, closed it, opened it again, and the page was lying on its side exactly as before. Nothing is broken. A viewer's rotate button turns the picture on your screen and throws that turn away when the window closes, because it never writes anything into the document. This tool writes into the document.
Where a page's orientation actually lives
Every page in a PDF carries its own rotation as a number stored inside the file, and every reader on the market obeys it. The rotate button in a viewer does not touch that number. Changing it is the entire job here, which is why the page comes back upright in Acrobat, in a browser, on a phone, in a print queue, and in whatever system you upload it to next. The turn travels with the file instead of living in one window on one machine.
Fix the one page that came out sideways
The usual problem is not a whole document. It is one page: the landscape table in the middle of a portrait report, the sheet the scanner grabbed the wrong way round, the fax page that arrived upside down. Drop a single PDF and every page appears as a preview. Each preview is drawn the way the page is really stored, so a sideways page looks sideways in the grid before you touch anything. Click the page that is wrong until it faces the right way, then download. Every other page comes back as it went in, with its own stored orientation unchanged.
The turn is relative, and that is what makes it work
Clicking rotate adds a quarter turn to whatever the page already has, the same way the rotate button behaves in every program you have used. This sounds like a footnote and it is the difference between a tool that works and one that looks broken. A page stored sideways at 270 degrees that is told to be 90 degrees does not move at all, which is exactly the page you were trying to fix. Here it composes instead.
| Page is stored at | You click | You get |
|---|---|---|
| 0, upright | 90 | 90, now landscape |
| 90, sideways | 90 | 180 |
| 270, sideways | 90 | 0, upright again |
| 180, upside down | 180 | 0, upright again |
Each click is 90 degrees to the right. Hold Shift while clicking to go left instead. The controls above the grid turn every page at once, and Reset puts the whole document back to how it arrived.
You see what will happen before it happens
The preview redraws as you click, so the page is upright in front of you before you commit to anything. Alongside it the tool states plainly how many pages will move and how many will not, and lists the turns grouped by angle, so a document where you touched pages 3 and 12 says so in one line. Rotation errors are the kind you do not notice yourself. Somebody else opens the document three days later and tells you. Being able to read back what you asked for is worth more here than on most tools.
Nothing is re-encoded
Turning a page means changing one number, so there is no decoding step and no re-rendering step. Text stays text and stays searchable, images keep the exact compressed bytes they arrived with, and fonts, links and bookmarks are untouched. On a ten page test with every page turned, the compressed page content came out at 13,014 bytes against 13,014 bytes going in, byte for byte identical. On a larger measured batch of three files and 46 pages, the input totalled 5,094,135 bytes and the output 5,093,217. Compare that with any tool that rotates by converting pages to images: that output comes back unsearchable, softer on paper, and often several times larger.
It also means the size of your document barely affects the speed. A 24 page report with 4 pages turned took 0.061 seconds of processing, and a 12 page file with every page turned took 0.016 seconds, because the work per page is one number either way. What takes time on a very large file is drawing the previews, and even that is quick: a 564 page PDF previews in about 10 seconds and a 1,631 page one in about 16.
A whole folder of scans at once
Add a second file and the tool switches by itself. The page grid is replaced by a list of files, each with its own rotation control, or one control that turns all of them together. Files you leave alone still travel through and come back with the rest, so a feeder that produced 40 upside down scans is one pass rather than 40 round trips. A single file downloads as a PDF under its own name; several come back as a ZIP with the original names kept, including Arabic and Chinese ones.
There is one deliberate limit in that mode: you rotate whole files, not individual pages. Page 12 does not mean the same thing in 40 different documents, and a page picker that pretended otherwise would turn the wrong page in 39 of them. For per page work, upload one file at a time.
Locked files, damaged files, and knowing it worked
A password protected PDF is accepted. Supply the password and the rotated copy comes back unlocked. A file damaged enough that it has to be repaired while being read is repaired, and the result says so in writing rather than handing you something quietly mangled; one that cannot be opened at all points you to Repair PDF instead. If you ask for the same page to turn two different ways, the request is refused and the page is named, rather than the tool guessing and turning the wrong way on the page you cared enough about to select twice.
After writing your file the tool reopens it, counts the pages again, and checks every page you selected against the orientation you asked for. A page that did not take the turn fails the job instead of reaching you as a file you believe is fixed. Your finished file is deleted from the server an hour after it is ready, and the job record is kept an hour longer so that a late download gets a clear message rather than a mystery.
What you get
- A rotation written into the PDF itself, so it survives closing, emailing and printing
- A preview grid that draws each page in its real stored orientation before you touch it
- Per page control on a single file, and whole file control on a batch
- Turns that add to what the page already has, so a sideways page composes back to upright
- Text, images, fonts and links returned unchanged, with nothing re-encoded
- No account, no watermark, and files deleted from the server an hour later
What's next?
Now that the pages face the right way, Compress PDF is the usual next step for a batch of scans. If the document needs cutting up rather than turning, Split PDF uses the same page picker, and Merge PDF puts a set of separate files back together.