PDF Split cuts a multi-page PDF into separate PDF files according to rules you give it. All of the work happens in the browser; the source file is neither uploaded nor modified.

The need usually sounds like “one document, several destinations”: a contract whose signature page must go to the other party on its own; a scanner that turned a stack of forms into one 100-page file when every 4 pages belong to one person; a journal issue whose PDF you want to strip down to your own article; or the opposite — a platform that caps the page count per file, so a long document has to go up in pieces.

Four ways to cut: by range (one file per range, or all ranges combined into one), every N pages (equal parts), selected pages (ticked in the thumbnails and pulled out as a new file) and page by page (one file per page). More than one output file is packed into a zip automatically; a single file downloads as a PDF.

The split works on the “current arrangement”, not the original file: delete a blank page or straighten a rotated one in the middle area first, and the page numbers and content follow the tidied state. That makes “tidy, then split” a single workflow, but it also means the page numbers you type refer to what is on screen.

1. 01Add a PDFdrag one in or click to choose. The workspace shows the page count and page thumbnails; an encrypted file asks for the open password first (viewing only — splitting still needs “PDF Unlock”).
2. 02Optional: tidy the pages firstrotate, delete or drag to reorder in the middle area. The split counts pages in the tidied order.
3. 03Switch to the “Split” paneland pick a method: - By range: type something like 1-3,5,8-10 into “Page ranges”; each range becomes one file. Turn on “Merge all ranges into one file” to join the ranges into a single file in the order you entered them. - Every N pages: fill in “Pages per file”; the document is divided from page 1 in equal parts, and the last part may hold fewer than N pages. - Selected pages: click pages in the middle area (Shift-click for a run); the panel shows how many are selected and the output is one PDF containing only those pages. - Page by page: one single-page PDF per page, all packed into a zip.
4. 04Click “Start splitting”the progress bar at the bottom advances page by page. Output names start from the source file name and carry a page label, for example contract_pages1-3.pdf; multiple files are packed as contract_split.zip.
5. 05Keep going (optional)“Download again” on the result card saves the files afresh; a single-PDF output can also be sent to “Add to workspace” for further work.

Two splits of a 12-page contract (contract.pdf, 11.5 KB) produce these real outputs:

The range syntax accepts these forms:

Forms that are rejected with a reason: 0-3 (pages start at 1), 5-2 (start page greater than end page), 10-15 (beyond the page count, the actual count is reported), a-b (not recognised).

By range: 1-3,5,8-
  contract_split.zip (10.2 KB), containing 3 files
    contract_pages1-3.pdf    3 pages
    contract_pages5.pdf      1 page
    contract_pages8-12.pdf   5 pages    ← "8-" means page 8 to the end
  pages 4, 6 and 7, which appear in no range, are not output

Every N pages: N = 4
  contract_split.zip (12.8 KB), containing 3 files
    contract_pages1-4.pdf
    contract_pages5-8.pdf
    contract_pages9-12.pdf

1-3,5,8-10     commas separate; a hyphen marks an inclusive range
1-3、5 8-10    the Chinese enumeration comma and spaces also separate
2—5  2~5       em dashes and tildes are normalised to "-"
9-             through to the last page
-3             from page 1 to page 3
1-3,2-6        ranges may overlap: pages 2 and 3 appear in both outputs

What "page 5" means inside a PDF

A PDF file has no “page 5” field. A page tree hangs beneath the document catalogue, with /Page objects as its leaves; a reader walks the tree depth-first, and the fifth leaf it meets is called page 5. That ordinal is the physical page number, and it has no necessary connection to a “5” printed on the page or to “Chapter 5 … 5” in a table of contents.

PDFs can also define page labels at document level, letting a reader show a preface as i, ii, iii and restart the body at 1. The range input here uses physical page numbers — the digits under each thumbnail. If your file has a Roman-numeral preface, enter the thumbnail numbers, not the printed page numbers.

How the split is carried out

Splitting and merging share one mechanism: pdf-lib creates an empty document, and pages are “copied” out of the source in groups. When a page is copied, the content streams, fonts, images and page-level annotations it references are deep-copied and renumbered along with it; page rotation is written through /Rotate. The groups are independent of one another — the output of group 1 knows nothing about group 2.

The “every N pages” grouping cuts after every N pages starting from page 1 of the current arrangement, and the last group takes the remainder; “by range” creates groups strictly in the order you enter, and overlapping ranges copy pages more than once; “selected pages” places the selected pages in one group in their arrangement order; “page by page” is simply N = 1.

Why the parts don't add up to the original size

Each extracted file must open on its own, so resources shared by several pages in the source — most typically embedded fonts — exist once in every output file. A 12-page document with 3 embedded fonts split into 3 parts keeps the font data 3 times. On top of that, the workspace copies pages one at a time to support per-page tidying, so a font shared by several pages of the same output is copied page by page as well (the “How it works” section of “PDF Merge” has the details). In the example above the source file uses PDF's built-in standard fonts and carries no embedded font data, so the three parts together stay about the original size; for a document with embedded fonts, the sum will be noticeably larger.

The reverse is also true: splitting does not “unlock” anything. If a page of the source holds a 20 MB scanned image, the extracted page is 20 MB.

What does not carry into the output

As with merging, copying pages takes only page-level content. Document-level structures — bookmarks, form field definitions, attachments, named destinations, document properties — do not enter any output file, and cross-page links that rely on named destinations stop working. A file with a digital signature loses the signature when split — the common result of every operation that rewrites a PDF's structure, not something particular to splitting.

Where the zip comes from

Multiple output files are packed into one zip by JSZip inside the browser (DEFLATE compression level 6). Streams inside a PDF are usually compressed already, so the zip almost never shrinks the total; its job is to let the browser download several files at once. Once unzipped, every PDF is complete and independent.

• Tidy first, then split: double-sided scans love to include blank backs. Tick the blank pages in the middle area, delete them, and then use “every N pages” — the count follows the post-deletion arrangement, and each part comes out even.
• Merging ranges = ordered extraction: with “Merge all ranges into one file” on, 8-10,1-3 produces one file holding pages 8–10 first and pages 1–3 after. That saves a download compared with split-then-merge, and it doubles as a quick way to reorder pages.
• Put the cover into every part: ranges may overlap. 1-4,1,5-8,1,9-12 yields 5 files — three 4-page sections and two lone covers; if what you want is “every section carries a cover”, the more direct route is one run of “merge into one file” per section with 1,5-8.
• Forms you hand out one by one: a single scan of 25 four-page enrolment forms split with “every N pages”, N = 4, gives 25 files whose names carry the page labels — easy to rename against the original numbering.
• Extract and keep working: “selected pages” outputs a single PDF that can go straight to “Add to workspace”, where a watermark or page numbers can follow without uploading again.
• Cut an oversized file into uploadable pieces: a 300-page scan headed for a platform capped at 50 pages splits into 6 parts with N = 50; if the cap is on size rather than pages, “PDF Compress” first usually pays off more.

Input format

PDF, one file at a time — the file currently selected on the left. Other files already added are unaffected.

Encrypted files

remove the password with “PDF Unlock” first; entering the viewing password alone is not enough to split.

Page basis

the current arrangement in the middle area (1-based, physical page numbers), not the numbers printed on the pages and not the original order if you have rearranged anything.

Output

a single group downloads as a PDF; several groups are packed into a zip whose compression depends on the content (PDFs generally shrink very little).

File size

no hard ceiling, though beyond 200 MB the tool warns that it will be slow; browser memory is the limit, since splitting holds the source and all outputs at the same time.

Kept

page content, vector text, embedded fonts, images, page-level annotations, rotation.

Not kept

bookmarks, form definitions, attachments, named destinations, original document properties, the validity of digital signatures.

What it does not do

it does not split by bookmark or chapter title, does not read the text to decide where to cut, and does not combine several source files (that is “PDF Merge”).

• "Entering 5-2 reverses the output": it does not. The start page must be less than or equal to the end page, or an error is reported. For reverse order, drag the pages into place in the middle area first.
• "The page numbers are the ones printed on the page": they are not. Use the ordinals under the thumbnails — especially important for a file whose preface uses Roman numerals.
• "Pages not written into a range are bundled into a file": they are not. Range mode outputs exactly the ranges you wrote; a page you forgot appears in no output.
• "The split files should add up to the same size": every output must work on its own, and shared fonts are duplicated, so the total is usually larger.
• "Splitting can strip passwords or signatures": an encrypted file cannot be split at all, and a signed file loses the signature because the document has been rewritten, not because the signature was “taken out”.

Files are read, parsed and rebuilt inside the browser and never pass through a server. Thumbnails are rendered locally by pdf.js, outputs are generated locally by pdf-lib, and the zip is packed locally as well. Everything still works offline, and closing the page frees the data in memory.

• ISO 32000-1:2008, PDF 1.7, §7.7.3 Page Tree (the definition of physical page order), §12.4.2 Page Labels (page labels and displayed page numbers): https://opensource.adobe.com/dcs/pdf/PDF32000_2008.pdf（访问日期：2026-09-08）
• pdf-lib documentation, copyPages and the document creation API: https://pdf-lib.js.org/（访问日期：2026-09-08）
• PKWARE, .ZIP File Format Specification (APPNOTE), the structure of a zip container: https://pkware.cachefly.net/webdocs/casestudies/APPNOTE.TXT（访问日期：2026-09-08）
• RFC 1951, DEFLATE Compressed Data Format Specification: https://www.rfc-editor.org/rfc/rfc1951（访问日期：2026-09-08）