BYTETOOLS

Auto-Rotate PDF

Detects the text direction on every page and rotates each one the right way up, with a confidence score and a manual override per page.

Drop a PDF here or click to browseReads the text direction on every page to find the right way up

What is the Auto-Rotate PDF?

Feed a stack of paper through a scanner and some of it comes back sideways or upside down, depending on how it went in.

  • Reads real text direction rather than guessing from page dimensions
  • Character-weighted voting so a sideways stamp cannot outvote the body text
  • Confidence score for every page
  • Per-page manual override with 0°, 90°, 180° and 270° options
  • Warns when a page has no text to analyse, as with un-OCR'd scans
  • Rotation is lossless — the page content is not re-rendered

How to use the Auto-Rotate PDF

  1. 1

    Drop the PDF onto the upload area and wait while every page is analysed.

  2. 2

    Review the table: it shows each page's current rotation, the detected upright rotation and a confidence percentage.

  3. 3

    Leave rows set to Auto to accept the detection, or pick an explicit angle for any page you want to control yourself.

  4. 4

    Pages with no extractable text are flagged — set those manually if they need turning.

  5. 5

    Click Apply Rotation & Download to save the corrected PDF.

About the Auto-Rotate PDF

Feed a stack of paper through a scanner and some of it comes back sideways or upside down, depending on how it went in. Fixing that by hand means opening the file, finding the rotate command, and repeating it page by page while keeping track of which ones you have already done.

This tool reads the actual text on each page instead. Every run of text in a PDF carries a transform matrix that describes the direction it is drawn in, so the tool measures that direction, weights each run by how many characters it holds, and works out the rotation that puts the baselines horizontal and running left to right.

You get a table showing what each page is set to now, what was detected, and how confident the detection is. Anything the tool got wrong — or any page with no text to read, such as a photo — can be overridden by hand before you apply the change. Processing is entirely local.

Frequently asked questions

How does the tool know which way up a page should be?

Every text run in a PDF has a transform matrix giving the direction it is drawn. The tool measures that angle for each run, weights it by character count, and picks the rotation that makes the majority of the text read horizontally.

Why do some pages say "no text"?

Those pages have no extractable text — usually scans that have not been through OCR, or pages that are just a photograph. There is nothing for the detector to measure, so they keep their current rotation unless you set one yourself.

Does rotating a PDF reduce its quality?

No. Rotation is stored as a page attribute, so the text and images are untouched and the file stays fully searchable and selectable. Nothing is re-rendered or recompressed.

What if the detection gets a page wrong?

Change that row's dropdown from Auto to the angle you want before applying. The detection is shown alongside a confidence score precisely so you can spot the doubtful pages quickly.

Can it fix pages that are tilted by a few degrees?

No — this corrects whole quarter turns only, which is what scanners and cameras produce. A slight skew from a crooked scan needs a deskew tool instead.

Related tools