Free PDF Rotate Pages Online

How It Works

1. Select or drop a PDF file above.
2. Use the batch controls to rotate all pages at once, or rotate individual pages.
3. Pages are rotated in your browser · nothing is uploaded to any server.
4. Download your rotated PDF instantly.

When Do You Need PDF Rotation?

PDF pages sometimes arrive in the wrong orientation, scanned sideways, uploaded the wrong way up, or exported from a mixed-orientation source document with one or two pages rotated incorrectly. Fixing orientation matters for document management, archiving, OCR pipelines, and any case where the document will be read by humans. This tool lets you correct the orientation of individual pages or entire documents in one pass, without installing software and without sending the file anywhere.

Rotation Options

• 90° Clockwise · Rotate each page a quarter turn to the right.
• 90° Counter-Clockwise · Rotate each page a quarter turn to the left.
• 180° · Flip each page upside down.

Frequently Asked Questions

Can I rotate individual pages?

Yes. Each page in the list has its own rotation controls. You can rotate pages independently or use the batch controls to apply the same rotation to all pages at once.

Is there a file size limit?

PDF files can be up to 100 MB. Since processing happens in your browser, very large files may take a moment depending on your device.

Is my PDF uploaded to a server?

No. All rotation happens locally in your browser. Your PDF never leaves your device, making this completely private and secure.

Does rotation affect PDF quality?

No. Rotation is a metadata operation that doesn't re-encode or compress the PDF. Your document maintains 100% of its original quality.

Can I undo a rotation?

Yes. You can rotate a page back by selecting the opposite direction. Just re-upload if you've already downloaded and want to start over.

What rotation is, in PDF terms

PDF stores rotation as a single integer in each page's dictionary, under the key /Rotate. The legal values are 0, 90, 180, and 270 (degrees, clockwise), and the entry tells the viewer or printer the clockwise angle by which the page should be turned when displayed or printed. The underlying content stream, the text-drawing and image-drawing operators that make up the page, is never modified by a rotation operation. This is the entire mechanism. PDF rotation is metadata, not content.

The implication is that rotation is essentially free in storage terms. The output file is byte-for-byte identical to the input except for the modified /Rotate value(s) in the affected page dictionaries and the new cross-reference table entries that the library emits when it serialises. A 100 MB scan rotates to a 100 MB output; a 100 KB text PDF rotates to a 100 KB output. There is no re-compression, no rasterisation, and no possibility of font substitution. The visible content is pixel-identical to the source when both are rendered through the same viewer at the new rotation.

A short history of /Rotate

The /Rotate entry has been part of the PDF specification since version 1.3 in 1999. Earlier versions of PDF allowed the more general /Matrix entry on page objects to specify arbitrary affine transformations, but the freedom to specify non-orthogonal rotations created interoperability problems: different viewers handled tilted rotations differently, and printed output sometimes did not match what appeared on screen. PDF 1.3 narrowed the contract to the four orthogonal rotations (0, 90, 180, 270 degrees), and /Rotate became the universal mechanism for specifying page orientation.

ISO 32000-1 in 2008 and ISO 32000-2 in 2020 preserved the same definition unchanged. The orthogonality restriction is, in practice, never a limitation. The real-world workflows that need rotation, scanner correction, mobile camera capture, fax orientation, mixed-orientation documents, all involve quarter-turn rotations. Non-orthogonal rotation of a page (such as a 45-degree watermark) is better expressed by drawing the rotated content inside a Form XObject, a separate mechanism that the PDF specification still permits for specialist use cases.

How the rotation actually happens

The rotation in this tool runs through pdf-lib, the same JavaScript library that powers the merge, split, and compress tools on this site. First, the source PDF is read into the browser tab through the standard File API. Second, pdf-lib parses the cross-reference table and exposes a list of page objects. Third, for each page the user selected, the tool reads the current rotation, computes the new total modulo 360, and writes the result into the page dictionary's /Rotate entry. Fourth, the document is serialised to bytes through PDFDocument.save() and delivered to disk as a download.

Bulk mode applies the same rotation to every page in one pass. Per-page mode applies different rotations to specific pages. Both modes are essentially constant-time relative to file size: the heavy work is parsing and serialising the cross-reference table, not iterating the pages themselves. A 500-page PDF rotates in roughly the same wall-clock time as a 5-page PDF, the difference being the I/O cost of reading and writing the larger file. The rotation arithmetic itself is microseconds per page.

Why rotated dimensions look strange to other tools

Each PDF page carries a small set of geometric boxes: MediaBox (the physical paper rectangle), CropBox (the visible area, defaulting to MediaBox), and the print-production boxes (BleedBox, TrimBox, ArtBox). The /Rotate entry does not modify any of these boxes. The boxes are defined in unrotated user space; the viewer applies /Rotate as a transformation matrix at render time, rotating both the page content and the box rectangles by the specified angle. The practical consequence: a US-letter portrait page (612 by 792 points) with /Rotate 90 still has a MediaBox of (0, 0, 612, 792) in the PDF source, but the viewer displays it as a 792 by 612 landscape page.

Tools that read PDF dimensions need to combine the MediaBox with the /Rotate value to get the on-screen orientation. Tools that ignore /Rotate will report the dimensions wrong, which is why some PDF-to-image converters or page-thumbnail generators occasionally produce sideways output. If you hand a rotated PDF to a downstream tool and the output is sideways, the fix is to flatten the rotation into the content stream, baking the orientation in so that even rotation-blind tools handle it correctly. The PDF Flatten tool does exactly that.

Real-world workflows that drive rotation

• Mobile camera capture. Mobile PDF scanning apps (Adobe Scan, Microsoft Lens, Apple Notes, CamScanner) capture pages through the phone's camera and assemble them into a PDF. The orientation metadata is sometimes correct and sometimes not, and the user often does not notice the wrong orientation until the PDF is opened in a desktop viewer or sent as an email attachment. Bulk rotation is the right fix here: every page from a single scanning session usually needs the same correction.
• ADF scanner mishaps. Office scanners with automatic document feeders take a stack of pages and digitise them in one pass. Mixed orientations within the stack are the dominant failure mode: the operator loads some pages upside down, the duplex unit flips a page incorrectly, or the document itself mixes portrait and landscape pages (a contract with an attached wide spreadsheet, say). Per-page rotation handles this case: only the affected pages need attention.
• Fax-to-PDF conversions. Older fax-server software converts incoming fax pages into PDF and applies a default orientation that often disagrees with what the sender intended. Pages from a fax may arrive sideways or upside down, particularly if the sender's machine was configured for a different paper size. Government, legal, and healthcare workflows where fax remains in active use see this pattern often. Bulk rotation is the standard fix.
• Mixed-orientation exports. A report with the body in portrait and one or two landscape pages for wide tables is common. When the report is exported to PDF from Word, Google Docs, or LibreOffice, the landscape pages sometimes export with /Rotate 0 instead of /Rotate 90, producing a document that displays correctly inside the source application but looks wrong in third-party viewers. Per-page rotation fixes the landscape pages without touching the portrait ones.
• Print-to-PDF from non-standard applications. Saving a print job as PDF from a CAD package, accounting program, ERP system, or embedded device sometimes produces files with unexpected /Rotate values, because the driver was written for paper output rather than on-screen display. The fix is rotation to a canonical orientation, usually with the wider edge of the page horizontal.
• Translation and review workflows. Bilingual documents, language-test answer sheets, and translated certificates frequently arrive with one orientation for the original-language pages and a different orientation for the translation pages. Per-page rotation keeps each section legible in its intended orientation.

Common pitfalls and workarounds

• OCR pipelines that ignore /Rotate. Some optical-character-recognition systems extract the raw content stream from a PDF and run their recognition on that stream directly, without applying the page's rotation matrix. A page that displays correctly in a viewer but has /Rotate 90 may produce sideways text in the OCR output. Fix: run the rotated PDF through the PDF Flatten tool to bake the rotation into the content stream, then run OCR on the flattened copy.
• Form fields rotate inconsistently. The PDF specification has historically been ambiguous about whether widget annotations (the visual part of form fields) should follow the page's /Rotate value or maintain their original orientation. Different viewers behave differently: Acrobat applies the rotation, older pdf.js builds did not, modern pdf.js does in most cases. The visible result is that form fields on rotated pages sometimes appear sideways. Fix: flatten the form before rotating, or apply rotation only to pages without form fields.
• Digital signatures break. A digital signature is a cryptographic hash of a specific byte range of the input file. Rotating a page changes the /Rotate entry, which changes the dictionary's byte representation, which changes the document's bytes, which breaks the hash. The output PDF will display "signature invalid" or "signature could not be verified" in any compliant viewer. This is the cryptographically correct behaviour, not a tool bug. If the document needs to remain signed, the rotation has to happen before the signature is applied.
• Coordinate-bound bookmarks may drift. PDF bookmarks and link annotations can target a specific (x, y) coordinate on a page. Rotating the page does not transform these coordinates; the bookmark still points at the same position in the rotated user space, which may now be outside the visible area. Most viewers round the destination to the nearest visible position, but the rounding sometimes lands on the wrong spot. Bookmarks that reference an entire page (the common case) are unaffected.
• Skew is not the same as rotation. If your scan is tilted by 5 or 10 degrees (not aligned to a quarter turn), this tool cannot help: the PDF specification only allows multiples of 90 in the /Rotate entry. Skew correction requires rasterising the page, applying an arbitrary affine transformation, and re-emitting the rotated raster, which produces a larger file with quality loss. The right fix for skew is to re-scan with the document aligned, or to use a desktop tool with image-processing capabilities (Adobe Acrobat Pro's Enhance Scans, or qpdf with a custom matrix).
• Encrypted PDFs fail to load. A PDF with an open password cannot be parsed until the password is supplied. This tool does not support encrypted input. Workflow: use the PDF Unlock tool to remove the protection first, rotate the unlocked copy here, and optionally re-apply protection with the PDF Protect tool on the output.

Why the rotated file is the same size

The output PDF differs from the input by a few dozen bytes total: the changed /Rotate integer(s) plus the new cross-reference table entries that pdf-lib emits when it serialises the modified document. Everything else, the fonts, the images, the vector graphics, the content stream operators, is written byte-for-byte from the input. A 100 MB input produces a roughly 100 MB output. This is the operational advantage of metadata-based rotation: it is the cheapest possible PDF transformation in storage terms, and the only PDF operation that can be reversed exactly (rotate by 90, rotate by 270, and you are back to the original file modulo cross-reference table reordering).

Browser-only versus cloud rotation

The cloud PDF rotation tools that fill the top of search results (Smallpdf, ILovePDF, Adobe Acrobat Online, PDF24's web app, Sejda, PDF2Go) all upload your source PDF to their servers, perform the rotation server-side, and serve the rotated output as a download. Their privacy policies state that uploaded files are deleted within a few hours, but the files transit the operator's network, exist on their disks during processing, and pass through whatever logging the operator maintains for abuse detection. Free tiers typically cap the number of operations per day; paid tiers remove the cap.

This tool does not upload. Your PDF is read into the browser tab through the standard File API, parsed by pdf-lib in the same tab, and the output is written back to disk through the standard download API. The only network traffic during a rotation is the one-time CDN load of pdf-lib when the page first opens. You can verify it: open the browser's developer tools to the Network tab, run a rotation, and watch that no requests fire with your file content. The cost of the privacy guarantee is browser memory: very large PDFs (typically over 100 MB on phones, several hundred MB on desktops) may run out of JavaScript heap. For files that fit, the output is interchangeable with what the cloud tools produce; for files that do not, the cloud tools have an advantage.

More frequently asked questions

Is the rotation "permanent"?

"Permanent" is the wrong word for what users usually mean. Once the /Rotate entry is written and the file is saved, every compliant PDF viewer displays the page at the new orientation, every time. The visible effect is permanent. The underlying content stream is unchanged, which is the right behaviour for almost every workflow: the PDF stays small, the rotation is exact, and a future tool can undo it without quality loss. Users who want a "permanent" rotation that bakes the orientation into the content stream (so that rotation-blind tools also see the new orientation) should use the PDF Flatten tool after rotating.

Can I rotate by an angle other than 90 degrees?

No. The PDF specification only allows multiples of 90 in the /Rotate entry. Arbitrary-angle rotation requires re-rendering the page content through a different mechanism, which produces a larger file with measurable quality loss. The tools that offer "any angle" rotation rasterise the page first, rotate the raster, and re-emit it as an image; the output is a different category of PDF (image-only) and loses any selectable text or vector graphics. For genuine skew correction, the right approach is to re-scan with proper alignment, or to use a desktop tool that combines rasterisation with skew detection.

What happens to a digital signature when I rotate a signed PDF?

The signature is invalidated. A digital signature is a cryptographic hash of the file's exact byte range; rotating the page changes the dictionary entry, which changes the bytes, which breaks the hash. There is no way to preserve the signature across rotation. The output PDF still displays the content correctly, but the signature panel in Acrobat or any compliant viewer will say "signature invalid" or "document has been modified since signing." If the document must remain signed, the rotation has to happen before the signature is applied. This is the cryptographically correct behaviour, not a tool limitation.

Why does my PDF look correct in Chrome but sideways in Word's PDF preview?

Almost always a viewer bug rather than a problem with the file. Chrome uses pdf.js, which has supported the /Rotate entry correctly for many years. Word's PDF preview, OneDrive's preview, some older e-readers, and a handful of legacy enterprise viewers occasionally ignore /Rotate and display the page at its un-rotated orientation. The workaround is to flatten the rotation: pass the rotated PDF through the PDF Flatten tool, which bakes the rotation into the content stream so that even rotation-blind viewers display the correct orientation.

Can I rotate a PDF on my phone?

Yes, if the file fits in browser memory. Mobile browsers typically allow JavaScript to use a few hundred megabytes of heap per tab, which is enough for most PDFs but not for very large scans. As a rough guide: text-only PDFs up to several hundred MB rotate fine on phones; image-heavy scans larger than 50 to 100 MB may run out of memory and the tab may crash. If a file is too large, the workarounds are to rotate it on a desktop, to split it into smaller files first with the PDF Split tool and rotate each piece separately, or to compress it through the PDF Compress tool before rotating.

Does rotation work with PDF/A archival files?

Yes. PDF/A is a constrained subset of PDF that requires the file to remain self-contained and reproducible: embedded fonts, no external dependencies, no encryption. The /Rotate entry is explicitly allowed in PDF/A and rotation preserves PDF/A compliance, because the operation does not touch any of the PDF/A-specific constraints. The output of a rotation operation on a PDF/A-compliant input is itself PDF/A-compliant, modulo the cross-reference table reordering that any save operation produces.

Related Tools