Pages—
Turned—
Misoriented—
Size—
Every rotate tool can turn a page. None of them can tell you which pages need turning, so a 90-page scan with four misfed sheets still means scrolling through 90 thumbnails looking for them. This one measures the direction the text runs on each page and reports the ones that are wrong — which is not the same as looking at the page's shape, because a landscape page is a legitimate page while writing that goes top-to-bottom is not. One button straightens each of them with its own turn. The rotation is stored as a page property rather than by re-drawing anything, so nothing is re-rendered and no quality is lost.
Step by step
01Open the document
Drop it anywhere on the page. Each page is drawn as a thumbnail and measured as it is drawn — ink coverage and the direction its text runs.
02Let it find the crooked pages
The orientation check reports how many pages read the wrong way and whether each is at a quarter turn or upside down. Straighten fixes every one of them with the turn it individually needs.
03Or turn pages yourself
With nothing selected a turn applies to the whole document. Click pages in the grid — shift-click for a run, or use the odd and even chips — to turn only those. Turns add up.
04Download the straightened file
The grid previews are already showing the exported orientation. The check re-runs against the staged turns, so the counter reaching zero is the tool marking its own work.
A worked rotation
The sample here is a five-page scanned batch where three sheets went through the feeder wrong. Every page is A4 portrait; nothing about their dimensions says which three.
Given
What the check reads on each page
page 1 text at 0° upright no turn
page 2 text at 90° reads downward turn 270°
page 3 text at 0° upright no turn
page 4 text at 270° reads upward turn 90°
page 5 text at 180° upside down turn 180°
after straightening: 0 of 5 pages misorientedThree misfed pages needed three different turns, which is why 'rotate all pages 90°' — the only control most tools lead with — cannot solve this document at all: it would straighten one page and break the two that were already right. The direction of the text is the only signal available, since every page here is the same shape. Its one limitation is worth knowing up front: a scan with no text layer has no text to read, so a pure image PDF has to be straightened by eye. Running OCR over it first gives the check something to measure.
Before you rotate
Rotation is a property, not a redraw
A PDF page carries a /Rotate entry of 0, 90, 180, or 270, and the viewer applies it when drawing. Setting it changes four bytes rather than re-rendering the page, so text stays selectable, images keep their resolution, and the file barely changes size. A tool that rasterises the page to rotate it is doing damage for no reason.
Page shape does not tell you a page is wrong
The instinct is to flag landscape pages in a portrait document, but a landscape page is usually a wide table that belongs that way — and a misfed portrait scan stays portrait-shaped after the feeder turns it. The direction the text runs is the signal that actually distinguishes them.
Turns add up, which is what makes them undoable
Each turn is added to whatever the page already carried, so two rights make a 180 and a right after a left is back where you started. That also means a page arriving at 90° and turned another 90° ends at 180°, not back at 90° — the arithmetic people expect and some tools get wrong.
A scan with no text layer cannot be checked
The orientation check reads the text layer, so a PDF that is nothing but page images has nothing for it to measure and reports no findings rather than guessing. Most scanners produce a searchable PDF with OCR text behind the image, which is enough; for one that does not, run OCR first or turn the pages by eye.
Rotation should not cost you the outline
Rotating moves no pages, so every bookmark still points where it did. Tools that rebuild the document to apply a rotation frequently drop the outline along the way, which turns a two-second fix into a lost navigation pane.
Rotating in a viewer is not rotating the file
The rotate button in most PDF readers turns the view and does not touch the document, so the next person to open it sees the original orientation. A permanent rotation writes the new value into the file, which is what these downloads contain.
The judgement call
Four situations, and what each one actually needs.
A handful of pages came out of the feeder sideways
Straighten
Each page gets the turn it individually needs, which a single document-wide rotation cannot do when the pages are wrong in different directions.
The whole document was scanned in landscape
Turn with nothing selected
One turn applies to every page, which is the right answer when every page is wrong in the same way.
Every second page is upside down
Even chip, then 180°
A double-sided scan with the wrong binding edge produces exactly this; selecting the even pages turns only those.
One wide table page is turned the wrong way
Click that page, then turn
The check will not flag it if its text runs correctly for its own orientation, so this is a judgement call the tool leaves to you.
An image-only scan with no text layer
By eye, or OCR first
There is no text to read the direction from, and the tool says so rather than reporting a confident guess.
Reference
FAQ
Rotate it in a tool that writes the new orientation into the file rather than only turning the view. The rotate button in most PDF readers changes what you see and leaves the document alone, so the next person to open it sees the original. The Tinapps rotate PDF tool sets each page's /Rotate value and hands you the rewritten file, which every viewer and printer then honours.
By the direction the text runs, not the shape of the page — a landscape page is often deliberate, while text running top-to-bottom never is. The Tinapps rotate PDF tool measures each page's text direction as it draws the thumbnails and reports how many pages read the wrong way, distinguishing a quarter turn from an upside-down page.
No, as long as the tool sets the page's rotation property rather than re-rendering the page. A PDF stores rotation as a /Rotate entry of 0, 90, 180, or 270 that the viewer applies while drawing, so nothing is rasterised: text stays selectable and images keep their resolution. The Tinapps rotate PDF tool works this way, and the output is within a percent of the original size.
Yes. In the Tinapps rotate PDF tool, clicking pages in the grid — or shift-clicking a run, or using the odd and even chips — limits a turn to those pages, and with nothing selected a turn applies to the whole document. Each page carries its own rotation in a PDF, so mixed orientations within one file are entirely normal.
Because a scan is a picture, and a picture of sideways text is not distinguishable from a picture of anything else without reading it. Orientation detection needs a text layer, which most scanners produce via OCR but some do not. The Tinapps rotate PDF tool reports that it found nothing rather than guessing, and running OCR over the file first gives it something to measure.
With most services, yes — the document is sent to a server and deleted on a schedule you have to trust. The Tinapps rotate PDF tool never sends it: the pages are drawn with pdf.js and rewritten with pdf-lib inside the page, so the file stays on your device and the tool keeps working with the network disconnected.
The turn is added to what the page already carries, so a page sitting at 90° that you turn another 90° ends at 180°. That is what makes rotations undoable — a left after a right puts the page back — and it is why the Tinapps rotate PDF tool shows the accumulated angle on each turned page rather than only the last turn applied.
It should not, because rotating moves no pages and every outline entry still points where it did. Some tools rebuild the document to apply the rotation and drop the outline on the way; the Tinapps rotate PDF tool carries the bookmarks across and reports how many survived.
Pages are drawn, measured, and rewritten entirely in your browser. The document is never uploaded, and the rotation is applied to a copy you download.
Keep going
Extract pages
Pull the straightened pages out once they are the right way up.
Split PDF
Cut the batch into per-document files after fixing the scan.
PDF tools
The rest of the category — merge, delete, and convert.
Image filters
For straightening and cleaning up a scan as a picture rather than a page.