Remove Duplicate PDF Pages
Review pages with matching geometry, text positions and color previews
About this tool
Find repeated PDF pages using conservative comparison before deciding which pages to remove. A suggested match needs the same page boxes, rotation, rendered dimensions, extracted text and text positions, plus the same color preview pixels. Pages with annotations are not suggested as duplicates. The first page in a matching group is the reference, and later matches are listed as candidates. Unlike approximate image similarity, this workflow does not claim to find shifted or noisy duplicate scans. A match is still only a review aid: previews are limited to 640 by 960 pixels and cannot prove that source content is identical. Different invoice amounts, layouts, colors or page geometry should remain distinct. No candidates are initially selected for removal. Review the original, select pages, keep at least one and confirm before exporting. The remaining source pages are copied to a new PDF, with a manual download and a 200 MiB / 80-page input limit.
Scenario Recipes
Remove an accidentally imported invoice copy
Goal: Keep distinct invoice amounts while removing a reviewed repeated page.
- Open the merged PDF and read the candidate links to earlier reference pages.
- Compare each candidate with its reference in the original PDF. Pay attention to amounts, dates, signatures and small scan details, even if the thumbnail matches.
- Select only verified copies, confirm the review and generate. Download the output and check that the distinct invoices remain in order.
Result: A new PDF with the selected later copies removed and different records retained.
Choose a later copy as the retained page
Goal: Control which representative remains when several pages match.
- Use the candidate label to identify the earliest reference page and its later matches.
- Keep the later page unselected and manually select the earlier page for removal. The checkboxes control removal; the candidate label is only a suggestion.
- Leave at least one page, confirm and generate. Check source-page numbers in the result list before downloading.
Result: The retained copy follows your explicit page selection rather than an automatic first-copy policy.
Failure Clinic (Common Pitfalls)
A repeated scan is not suggested
Cause: Noise, shifts, extracted-text differences or geometry changes prevent exact preview matching.
Fix: Use the original to review both scans manually. There is no approximate similarity setting; increasing a tolerance is not part of this tool.
Matching previews conceal a meaningful source difference
Cause: Finite preview resolution and renderer support cannot establish full document equivalence.
Fix: Treat a match as a candidate. Verify small print and scan details in the source before selecting it for removal.
The same text in different positions is kept
Cause: Text positions, page geometry and rendered colors participate in the comparison.
Fix: This is deliberate. If the layout difference is irrelevant to your task, inspect and select the unwanted page manually.
Pages with annotations never match automatically
Cause: Comments or links may carry information not safely represented by a visual comparison.
Fix: Review the original annotations separately. If you copy annotated pages, check the downloaded document; cross-page link destinations are not guaranteed.
Production Snippets
Similar invoice layouts can contain different business data
text
Page 1: all white
Page 2: all black
Page 3: all red
Page 4: all blue
Page 5: Invoice total $100.00
Page 6: Invoice total $999.00
Page 7: same content, position and page geometry as page 5
Page 8: $100.00 invoice moved to a different position
Page 9: all white, landscape geometry
Suggested duplicate: page 7 of page 5 only.
Initially selected for removal: none.Frequently Asked Questions
How are duplicate candidates compared?
Geometry, extracted text and its positions, and full RGBA pixels from a color preview must match. Matching uses SHA-256 fingerprints held only in memory. It does not rely on a coarse perceptual similarity threshold.
Does a match prove identical source content?
No. The comparison preview is limited to 640 Γ 960 pixels and may miss details or unsupported features. Open the original PDF and verify each proposed duplicate before removal.
Can it remove shifted scans or near-duplicates?
There is no approximate or sliding-window mode. The comparison covers all accepted pages, but visual noise, shifts, text differences and page geometry can prevent a suggestion. Use manual review if two scans differ.
Which copy is retained?
The first page of a matching group is the reference; selecting candidates chooses later matching pages only. All pages start unselected, and manual checkboxes let you choose a different page. At least one page must remain.
What PDF files can I use?
Choose one unencrypted PDF from 1 to 80 actual pages and up to 200 MiB. Forms, XFA, signature permissions and portfolios are rejected. Rendering also has page-dimension and pixel budgets; split large documents first.
What survives in the output?
Selected source pages are copied in their original order, including text and vector content. The output is a new PDF: document-level outlines, attachments, forms, signatures and working cross-page destinations are not preserved or guaranteed. Review each downloaded file.
Is the PDF uploaded or downloaded automatically?
The document is processed in your browser and is not uploaded. Generated files stay in the result list until you download them individually. Changing the source, selection or detection setting clears old output.
Keep browsing