How to correct page orientation across a multi-page scanned pdf
- Step 1Audit every page and group by needed angle — Page through the scan and build three lists: pages needing
90° clockwise, pages needing270°(counter-clockwise), and pages needing180°. This grouping maps directly to the passes you'll run. There's no thumbnail picker — you select by typing numbers. - Step 2Open the Rotate tool and load the scan — Drop the multi-page scan into the PDF Rotate tool. The page count shows next to the filename — check it against your viewer so your page numbers line up.
- Step 3If the whole scan is wrong the same way, do one blanket pass — Pick the single correct angle and leave Pages as
all. One Process run turns every page. Skip the remaining steps if this covers everything. - Step 4Otherwise, run one pass per angle group — Process the
90°group (type those page numbers), download, then re-load and Process the270°group, and again for the180°group. Keep the page lists disjoint so additive rotation doesn't stack on a page twice. - Step 5Type page numbers comma-separated — Use 1-based numbers like
3, 7, 12, 18. Hyphen ranges aren't supported, so on a long scan you must list each page — a spreadsheet column of numbers pasted as comma-separated text helps. - Step 6Download and re-verify the whole scan — Open the final PDF and scroll the entire document to confirm every page is upright. If you'll OCR the scan, do it now — orientation must be correct first for accurate recognition.
Two modes for a multi-page scan
Use a blanket turn when the whole scan is wrong identically; use named-page passes when orientations are mixed. One angle per Process run either way.
| Scan situation | Pages value | Passes needed |
|---|---|---|
| Every page wrong the same way | all | 1 (single angle) |
| Some pages 90° CW, rest fine | list the 90° pages | 1 |
| Mix of 90° CW and 270° pages | list each group separately | 2 (one per angle) |
| Mix of 90°, 270° and 180° | list each group separately | 3 (one per angle) |
Page-list rules that bite on long scans
On a 50-page scan these matter more than on a 3-page file. The Pages field is comma-split and integer-parsed, with no range support.
| Situation | What happens | Do this instead |
|---|---|---|
You list 3-18 to cover a run of pages | Reads as just page 3 | List each page: 3, 4, 5, …, 18 |
| Same page named in two passes | Angles stack additively (e.g. 90°+90° = 180°) | Keep page lists disjoint across passes |
| A page number > total pages | Job errors | Stay within 1 to the page count shown |
| Non-numeric or empty tokens | Silently dropped | Fine — clean numbers still apply |
Free vs Pro limits for long scans
From the JAD Apps tier table. A 50-page scan is exactly the free page ceiling; high-DPI scans hit the size cap first.
| Limit | Free | Pro |
|---|---|---|
| Max file size per job | 2 MB | 50 MB |
| Max pages per PDF | 50 | 500 |
| Files per job | 1 | 5 |
Cookbook
Multi-page scan recipes. The output page count always equals the input — rotating never adds or removes pages, so use it as a quick sanity check.
Whole ADF scan came out sideways
A batch of landscape originals fed into a portrait scanner produced 30 sideways pages, all the same way (top edge on the left). One blanket pass fixes them all.
Rotation angle: 90° clockwise Pages: all After: all 30 pages upright; output still 30 pages
Mixed scan needing two passes
In a 24-page scan, pages 5, 6, 7 are sideways one way (90° CW) and pages 15, 16 are sideways the other way (270°). Two passes, disjoint page lists.
Pass 1: 90° clockwise Pages: 5, 6, 7 download corrected file Pass 2: 270° counter-clock. Pages: 15, 16 After: all 24 pages upright
Three-angle scan handled in three passes
A messy scan has clockwise, counter-clockwise, AND upside-down pages. Group by angle and run one pass each, never reusing a page number.
Pass 1: 90° clockwise Pages: 2, 9 Pass 2: 270° counter-clock. Pages: 4, 11 Pass 3: 180° Pages: 6 Lists are disjoint, so no page is rotated twice.
Avoid double-rotating a page across passes
If you accidentally name page 5 in both the 90° pass and a later pass, its rotations stack. Keep lists disjoint, or restart from the original.
Pass 1: 90° CW Pages: 5, 6, 7 Pass 2: 90° CW Pages: 5, 8 ← page 5 repeated! Page 5 now has 90+90 = 180° (upside down). Fix: re-load original; run with disjoint lists.
Rotate the scan, then OCR it
OCR accuracy depends on upright pages. Correct every page's orientation first, then run OCR so the searchable text layer matches the now-upright scan.
Step 1 Rotate all misoriented pages (one or more passes) Step 2 Download the upright scan Step 3 Open /pdf-tools/pdf-ocr and run OCR Result: accurate, aligned searchable text across the scan
Edge cases and what actually happens
You tried a hyphen range to cover many pages
By designOn a long scan it's tempting to type 3-18, but the Pages field reads it as just page 3. You must list each page comma-separated. Pasting a comma-separated number sequence (e.g. from a spreadsheet) is the practical workaround. Range syntax exists in delete-pages and extract-pages, not in Rotate.
A page got rotated twice across two passes
DoubledRotation is additive, so naming a page in more than one pass stacks the angles. Keep the page lists for each pass disjoint. If a page ended up doubled, re-load the original scan and re-run with non-overlapping lists.
Long scan exceeds the 50-page free limit
Free limitA scan over 50 pages is blocked on the free tier with a page-count notice. Pro raises the cap to 500 pages. To stay on free, split the scan into ≤50-page chunks with the Split tool, correct each, then re-merge with Merge.
High-DPI scan exceeds the 2 MB free limit
Free limitEven a short scan at high resolution can pass 2 MB and be blocked on free. Pro raises the size cap to 50 MB. To stay on free, reduce the scan size first with Compress (lossy).
A page number is beyond the document length
errorNaming a page that doesn't exist makes the job fail. On a long scan it's easy to mistype — the page count shown after upload is the valid range (1 to N). Double-check your highest number.
OCR was run before fixing orientation
Re-run OCRIf the scan was OCR'd while pages were misoriented, the recognised text is wrong even after you rotate the pages. Fix orientation first, then re-run OCR to regenerate an accurate, aligned text layer.
Did the file get larger or slower?
PreservedNo — rotation only rewrites the orientation flag on the named pages; the page rasters are reused. The output size stays close to the input and viewers render it at normal speed.
Need each page auto-detected and straightened
Out of scopeThis tool does not auto-detect orientation; you specify the angle and pages yourself. There's no content-aware deskew or auto-rotate. For automatic orientation detection you'd need an OCR-based deskew step outside this tool — but for known angles, the manual passes here are exact and lossless.
Frequently asked questions
Can I rotate different pages by different amounts in one operation?
No — each Process run applies one angle to the pages you name. For a mixed scan, run one pass per angle group: a 90° clockwise pass, a 270° pass, and a 180° pass as needed. Keep the page lists disjoint so additive rotation doesn't stack on any page twice.
How do I fix a whole scan that's wrong the same way?
Pick the single correct angle and leave the Pages field as all. One Process run turns every page. This is the fastest path when an entire ADF batch came out sideways or upside down identically.
Why did only the first page in my range get rotated?
The Pages field doesn't support hyphen ranges. 3-18 is read as just page 3. List every page comma-separated. For long runs, paste a comma-separated number sequence from a spreadsheet.
Will rotating slow down my viewer or bloat the file?
No. Rotation is a lightweight page-level change — only the orientation flag is rewritten and the page rasters are reused. The output file size stays close to the input and renders at normal speed.
Should I rotate before or after running OCR?
Rotate first, then OCR. OCR accuracy depends heavily on upright pages — recognising sideways or upside-down text is unreliable. Correct every page's orientation, then run OCR so the searchable text matches the upright scan.
Does the tool detect which pages are misoriented automatically?
No. You identify the misoriented pages and choose the angle. There's no auto-detect or content-aware deskew — but for known angles, manual passes are exact and don't degrade the scan.
My scan has more than 50 pages — can I still fix it?
On the free tier, no — PDFs are capped at 50 pages. Pro raises the cap to 500 pages. To stay on free, split the scan into ≤50-page chunks with the Split tool, correct each, then re-merge with Merge.
What if I accidentally rotate a page twice?
Rotation is additive, so the angles stack (e.g. 90°+90° = 180°). Keep each pass's page list disjoint to avoid it. If it happened, re-load the original scan and re-run with non-overlapping lists — that's cleaner than trying to subtract a rotation.
Is my scanned document uploaded anywhere?
No. Processing runs in your browser via pdf-lib, so scanned records — contracts, statements, case files — never leave your device. Only an anonymous usage counter is recorded when you're signed in, with no file content.
Will this keep my scan archive-quality?
Yes. The original page images are reused untouched; only the orientation entry changes. There's no re-rasterisation, so the corrected scan is as faithful as the input — suitable for archival storage.
Can I batch-process several scans at once?
Rotate handles one file per job (Pro allows up to five files per job). For multiple separate scans, process them in sequence — each download is independent. Within a single scan, you can fix any number of pages.
How do I select pages without thumbnails?
You type the page numbers. The tool shows the document's page count next to the filename, and you select by entering 1-based numbers (comma-separated) in the Pages field. There's no thumbnail grid or click-to-select — note the page numbers from your viewer first.
Privacy first
All PDF processing runs locally in your browser using PDF-lib and pdf.js. No file is ever uploaded — only metadata counters are saved for signed-in dashboard stats.