It works out which pages are sideways for you. PDF records the angle of every line of text, so the pages that need turning can be found rather than hunted for. A scan where page 2 is on its side and page 7 is upside down is one click, not a page-by-page inspection.

Every other rotator turns the whole document by one angle — which fixes half a mixed-up scan and breaks the other half. Nothing is uploaded and nothing is re-encoded: rotation in PDF is a property of the page, so the result is instant and pixel-for-pixel identical to the original.

Rotate PDF

Local

Choose a PDF, or drop one here Nothing leaves this tab

Rendered by PDF.js and written by pdf-lib, both fetched when you add a file. Your document is not part of that request.

How sideways pages are found

A PDF does not store a picture of a page. It stores instructions, and every line of text carries a transform matrix saying where it goes and which way it faces. The angle is the first thing in that matrix:

The angle of a line of text

matrix = [a, b, c, d, e, f]
angle = atan2(b, a)

Dominant text angle	What it means	Correction applied
0°	Reads normally	None
90°	On its side, anticlockwise	Rotate 270°
180°	Upside down	Rotate 180°
270°	On its side, clockwise	Rotate 90°

The angle is taken per page and weighted by how much text sits at each one, so a sideways caption on an otherwise upright page does not throw the result. Pages the detector believes are wrong are marked in the grid, and you can accept the lot or override any of them.

Scanned pages carry no text, so nothing can be detected on them. A photographed or scanned page is an image; there are no text runs to read an angle from. Those pages are left alone and you rotate them by eye — which is what every tool requires, this one simply says so rather than guessing.

Rotation is metadata, not pixels

Each page in a PDF has a /Rotate entry holding 0, 90, 180 or 270. Changing it tells every viewer to display the page turned; the content underneath is untouched:

	Rotating a PDF	Rotating an image
What changes	One number per page	Every pixel is rewritten
Quality	Identical — nothing is re-encoded	Lossy formats degrade a little each time
File size	Unchanged	Varies
Speed	Instant, even on 300 pages	Proportional to the pixels
Text	Still selectable and searchable	Not applicable

This is why rotating a PDF is one of the few edits that costs nothing at all. It also means a viewer that ignores the /Rotate entry — a handful of very old ones do — will show the page unrotated. Every current reader, browser and phone honours it.

Which pages actually need turning

Situation	What to do
A whole document sideways	Use All left or All right. Usually a scanner fed the pages the wrong way
A mixed-up scan	Accept the suggestion. This is the case other tools cannot handle
Landscape tables in a portrait report	Rotate only those pages, so they read without turning the screen
Alternating pages from a duplex scanner	The detector catches this exactly — every second page at 180°
A page that is genuinely landscape	Leave it. Wide content on a wide page is correct, not a defect

Common mistakes

Rotating the whole document to fix two pages. It is the only thing most tools offer, and it turns a document with two problems into a document with twenty. Check which pages are actually wrong before applying anything to all of them.

Rotating in a viewer and expecting it to save. Most PDF readers let you turn the view without changing the file — the next person to open it sees the original. Some offer a separate save; many do not make the difference obvious. A rotation only sticks if the file is written again.

Converting to images to rotate them. That replaces a free metadata change with a lossy re-encode, loses every bit of selectable text, and multiplies the file size. Rotation is the one edit that costs nothing — keep it that way.

Uploading a scanned contract to turn one page. Rotation is metadata — it needs no server, no processing power and no account. A tool that asks you to upload for this is asking for more than the job requires.

Frequently asked questions

How does it know which pages are sideways?

Every line of text in a PDF carries a transform matrix, and the angle comes straight out of it. The dominant angle on each page — weighted by how much text sits at each — says whether the page reads normally, sits on its side or is upside down. Pages that look wrong are marked, and you can accept them all or override any one.

Does rotating lose quality?

Not at all. Rotation is a single number stored on each page telling viewers how to display it — nothing is re-encoded, the file size does not change, and text stays selectable. This is why it is instant even on a very long document.

Why did it not detect anything on my scan?

Because a scanned page is an image, with no text runs to read an angle from. Those pages are left alone and you turn them by eye using the buttons on each thumbnail. Detection needs a text layer, which a scan only has if it has been through OCR.

Is my file uploaded?

No. It is read and rewritten by your browser — the network panel in developer tools shows no request carrying your document. The only network requests are for the two libraries, and they happen once.

Can I rotate just one page?

Yes. Each thumbnail has its own rotate button, and the toolbar buttons apply to everything if you want the whole document turned. A landscape table inside a portrait report is exactly the case for rotating a single page.

Why does my rotation disappear when I reopen the file?

Because the viewer rotated the view rather than the file. Most readers let you turn a page on screen without changing anything on disk, and do not always make that obvious. A rotation only persists if the file is written again, which is what the save button here does.

Does the existing rotation get replaced or added to?

Added to. A page already stored at 90° that you turn another 90° ends at 180°, which is what you would expect from watching the thumbnail. The original value is respected rather than overwritten.

Can I rotate a password-protected PDF?

Not while it is protected — remove the password in a PDF reader first. The saved file is not encrypted, so protect it again afterwards if the original was.

Related tools

See the full list of Generators, or try:

Merge PDF — combine documents with page-level control
Split PDF — several ranges in one pass
Images to PDF — photos, the right way up
PDF to Images — pages as PNG or JPEG
PDF to Markdown — the text rather than the layout

Rotate PDF Free: Detects Sideways Pages Automatically