Pick a page range (or single pages) and download them as a new PDF. Processed locally — your file never leaves your computer.
Drop a PDF here, or click to choose
Encrypted (password-protected) PDFs are not supported.
Comma-separated ranges and single pages, 1-based.
Pull the pages you need out of a PDF and save them as a new file — a signed page, one chapter, a single invoice from a monthly bundle.
Under the hood, the range string is parsed into a list of page indices, and pdf-lib copies those page objects from the source document into a brand-new one — the same object-copying approach as the merge tool, so text stays selectable and vectors stay sharp. Descending ranges like 9-3 extract in reverse order, which is occasionally exactly what a scan needs.
Invalid page numbers are simply ignored rather than fatal: asking for pages 3-99 of a 10-page PDF extracts 3 through 10. If nothing valid remains, the tool says so instead of producing an empty file. Out-of-bounds checks use the actual page count reported when the file loaded, so the range hint always matches your document.
As with every tool here, the file is processed in browser memory only — court exhibits, medical records and financial statements can be split without a server ever seeing them.
Typical workflows this page replaces: sharing one invoice from a supplier's monthly bundle, sending a counterparty only their signed pages from a contract pack, or splitting a scanned textbook into chapters. Because extraction copies page objects, the output pages remain text-searchable — you can extract a page and still Ctrl+F it, which screenshot-and-crop approaches destroy.
Extraction also doubles as a repair tool. A PDF that crashes one particular viewer can sometimes be saved by extracting all its pages ('1-999' — extras are ignored) into a fresh document, leaving whatever structural quirk crashed the original behind. It won't fix every corrupt file, but it's the first thing to try before declaring a document lost.
Invert the range: to delete page 4 from a 10-page file, extract '1-3, 5-10'. The result is the original minus the unwanted page.
No — remove the password first (open with the password, save without security), then extract.
Only if you write the range in descending order: '9-3' produces pages 9 down to 3. Ascending keeps normal order.
Memory again — thousands of light pages work; very heavy scanned files may need splitting into two runs.
Yes — pages are extracted as-is, images included. What you lose on scans is searchability, which extraction neither creates nor destroys.
Yes — 1-based counting matching the viewer's page display, not the printed page labels (which can differ on books with roman-numeral front matter).
One file at a time here — extract from each, then use the Merge tool to combine the extracts in order.
QKitBox