Split PDF Online
Extract specific pages or ranges. Example: 1-3,5,8-10.
Drop a PDF here
or choose a file
The range string is the whole feature
Split PDF parses a list such as 1-3,14-62,150-168 into zero-based indexes, then pdf-lib copies those pages into a new document. It does not redraw them. A text page stays text. The script walks commas, then hyphens, and ignores junk tokens that do not start with a number. There is no byte cap on a server. One heavy file still has to sit in the tab, which is easier than a multi-file merge and still enough to kill a phone if the scan is near 100 MB.
The site is HTTPS. This operation never POSTs the PDF. Nothing is stored to wipe, because the Worker is not in the path.
| Behavior | |
|---|---|
| Engine | pdf-lib page copy |
| Input | One PDF. Encrypted files fail |
| Output | One PDF containing the indexes you listed |
| Size | Sum of the kept pages, not a compressed rewrite |
Cuts that match a real boundary
A reviewer who only marks one chapter
Send findings, not the map atlas. If that chapter is still over the portal limit, compress that extract and leave the split master next to it.
Signature pages glued to the section they close
A range that ends the clause before the signature block creates a file nobody can execute. Include the block. Write the ranges in English on a note first, then translate to numbers.
One page for a slide
Split that page out before PDF to JPG. Rendering every page at scale 2 to grab one image is how laptops sound like fans.
A question with a small context window
Chat and summarize behave better on the chapter that contains the table. The archive file stays intact for the appendix you did not need today.
Type the range, then open both ends
- Drop one unlocked PDF.
- Type viewer indexes, for example
14-62. Do not type the printed footer if a cover shifted it. - Extract. Open the download. The first and last pages should be the boundary you named.
- Rename with the range in the filename:
findings-p14-62.pdf.
Blank pages at a break might be intentional duplex backs. Look before you delete them in Organize. More on ranges: a 180-page report cut three ways.
Indexes the parser will skip
A token that does not start with a digit is ignored. pp. 4-9 may not mean pages 4 through 9, because the letters get in the way. Type 4-9. A range that runs backwards, or a number past the last page, should be checked in the output rather than trusted. The new file’s page count is the count of indexes that resolved, not the count of numbers you thought you typed.
Name the extract with the range. Six months later extract.pdf is useless next to three other extracts. The master stays the file you split from. If you need two non-contiguous spans as two attachments, run the page twice. One run produces one PDF.
Frequently asked questions
Which page numbers does the range box use?
The viewer index, starting at 1 for the first page in the file. Printed footers that say “iv” or restart at 1 after a cover are not what the parser reads. Check the thumbnail index before you type.
What syntax is accepted?
Commas separate items. A hyphen is inclusive. 1-3,8,10-12 is cover plus two later spans. A single number is one page. An empty field is not a plan. Type the ranges.
Will the extract get smaller in megabytes?
Only by leaving pages behind. A 12 MB map plate is still a 12 MB map plate if that page is in the range. Splitting does not recompress images.
Can I split a locked PDF?
Not until it opens. Load fails on encryption the same way merge does. Unlock a permitted copy first. Splitting is not a confidentiality control.
Is the original deleted?
No. You download a new PDF. Keep the master. Extracts are working copies, and six weeks later someone will ask for a page you did not extract.
Why would I split before chat or OCR?
The model only sees the file you give it. One chapter is a better question than a 180-page dump, and you do not send pages you never meant to transmit.