Fix sideways scans and upside-down pages.
Add a PDF, pick a rotation, and choose whether it applies to every page or just the ones you name. Rotation is stored as page metadata rather than redrawn, so nothing loses quality.
Add the PDF. The panel reports the page count, which is the range your page numbers have to stay inside.
Ninety degrees clockwise fixes a page scanned sideways with the top to the left. Two hundred and seventy fixes the opposite. One hundred and eighty is for pages fed upside down, which happens with duplex scanners more often than anyone admits.
Leave the page field blank and every page rotates. Fill it in with the same range syntax as the split tool to rotate only some — useful when a document has one landscape table in the middle of a portrait report.
The rotation is added to whatever the page already had, so applying ninety twice gives you a hundred and eighty. Open the result and confirm before sending, because a page that looked wrong in one viewer occasionally looked right in another.
Every page object in a PDF carries a Rotate entry: an integer that must be a multiple of ninety. It tells the viewer how far to turn the page clockwise before displaying it. Setting it does not touch the drawing instructions, the embedded fonts, or the images. This is why rotation here is instant and lossless — the tool changes one number per page and saves the file.
The alternative implementation, used by some tools, is to rasterise the page, turn the resulting bitmap, and write it back as an image. That produces a file that looks right and is much worse: text is no longer selectable or searchable, file size balloons, and print quality drops to whatever resolution was chosen. If a rotate tool takes noticeable time on a small file, it is probably doing this. Check whether you can still select text in the output.
Because the entry is absolute, applying rotation twice adds up. A page already at ninety, rotated ninety again, is stored as a hundred and eighty. The tool reads the existing value and adds to it rather than overwriting, which is what you want when a document has mixed orientations already.
A PDF page defines several rectangles. The MediaBox is the full sheet. The CropBox is the region a viewer displays, which is how a page can appear trimmed while still containing content beyond the visible edge. There are also BleedBox, TrimBox and ArtBox, used in print production. Rotation is separate from all of them: it is an instruction to turn whatever those boxes describe, applied at display time. This is why rotating never crops anything and never moves content relative to the page — the geometry is untouched and only the viewing angle changes. It is also why a page that looks wrong because its CropBox is set oddly cannot be fixed by rotating it; that is a different problem needing a different operation.
Rendering a PDF page is expensive: the viewer must parse the content stream, resolve font references, decode embedded images, and rasterise the result. Doing that on every scroll would be unusable, so viewers render a page once and keep the bitmap. The cache is usually keyed on the file path and the page number rather than on the file contents, so overwriting the file on disk while it is open leaves the viewer showing a bitmap generated from the previous version. Nothing is wrong with your file; you are looking at a picture of the old one. Closing the document discards the cache. Browsers are the most stubborn case because the file is also in the HTTP cache, which is why a hard refresh is sometimes needed on top of reopening the tab.
Viewers cache aggressively. A file open in a reader while you overwrite it on disk will often keep showing the old render. Close the document and reopen it. Browser tabs are the worst offender here — a hard refresh is sometimes needed. If the rotation is genuinely absent, the file may have been written by a tool that stores orientation in the media box instead, which is unusual but does happen with old scanning software.
Close and reopen the file first, because viewers cache renders. If it persists, try rotating the other way — a page that needs 270 looks identical in a thumbnail to one that needs 90, and it is easy to guess wrong.
The page field was left blank, which rotates the whole document. Rotate everything back by applying the complementary angle, then run it again with the specific pages listed.
Print dialogs have their own orientation and scaling settings that can override the document. Check that the print dialog is set to portrait and to actual size rather than fit-to-page.
It needed 270 rather than 90. Apply 180 to the same page to flip it the rest of the way.
The file is rewritten when saved, which reorganises the object structure. A small difference either way is normal and nothing has been lost.
Most PDF readers can rotate the view with a keyboard shortcut, but many only rotate what is on your screen and do not save it into the file. The recipient opens it and sees the original orientation. Check that your reader offers a save, not just a view rotation. Preview on macOS saves properly; several free Windows readers do not.
If the scan is also crooked, faint, or cropped, rotating it is treating a symptom. Feeding the page through the scanner again correctly takes about the same time and fixes everything at once.
They perform the same one-number change and require your document to travel to a server to do it. For a task this trivial the privacy trade is particularly poor value.
No. A PDF page carries a rotation value, and this changes that value. The page content is not re-rendered.
Yes. Leave the page field blank to rotate everything, or list pages and ranges to limit it.
Some viewers cache the previous version. Close and reopen the file, or refresh the tab.