Extract Single Pages From a PDF

Pick the mode first: Ranges for one file of selected pages (1-3, 5, 12-), Every page for one file per page, Chunks for fixed-size parts. Each output is a new PDF containing only copied pages — unselected pages are genuinely absent — and the whole split runs in your browser, so nothing is uploaded.

100 % local

Vous voulez juste l’outil ? Ouvrez Diviser un PDFExtrait des pages et laisse le reste intactOuvrir l’outil

Le guide ci-dessous n’est disponible qu’en anglais.

Splitting is what you do when you need to send one page of something: the signed page of a contract, a single invoice from a scanned batch, the medical letter but not the rest of the file. Which is to say, splitting is what you do to documents you are being careful with — and most online splitters ask you to upload the entire document first, to a server whose only role was to move it back to you.

Here the split runs locally, in a Web Worker, using pdf-lib. Each output is not a copy of the original with pages hidden — it is a new document into which the selected pages are copied with their complete dependency graphs. Nothing from the pages you did not select exists in the output, which is the difference between extracting page 3 of a contract and sending a file that merely displays page 3.

This page covers the three split modes and the jobs each fits, the range syntax (deliberately the same one print dialogues use), how the output files are named, and the honest answer on bookmarks and metadata — what carries into the parts, what stays with the original, and why.

All three modes live in the Split PDF tool — drop one PDF, choose a mode, and set the output naming pattern.

For why a lazy split leaks the pages it claims to remove, the guide to PDF structure walks through page trees, content streams and cross-reference tables.

Three modes, three jobs

The mode selector matches the reason you are splitting in the first place:

Ranges — one output file containing the pages you named, in the order you wrote them. The mode for pulling out the part you need.
Every page — one output per page, as many files as the document has. The mode for a scanned batch that should have been separate documents: forty invoices scanned in one pass.
Chunks — sequential groups of N pages, the last one short. The mode for a size limit somewhere else: a 10 MB upload gate, an email attachment ceiling, a print shop that takes twenty pages per job.

Range syntax

The order-as-written rule has real uses: the output follows the sequence you typed rather than a sorted one, so 5, 1, 3 produces a document in that order, and a duplicate page number is occasionally exactly what you want when a cover sheet has to appear twice. If reordering is the whole job rather than a side effect of selecting, the page reorder tool is the more direct instrument.

You type	You get	Note
1-3, 5	Pages 1, 2, 3 and 5	Commas separate the parts
12-	Page 12 to the end	Open end runs to the last page
-5	The first five pages	Open start works too
5, 1, 3	Those pages, in that order	Order as written; duplicates allowed

Naming the outputs

A split that produces one file per page cannot give every file the same name, so outputs follow a pattern: the document's own name plus a counter — report-1.pdf, report-2.pdf and so on. The pattern is editable, with four tokens that cover most jobs:

{name} — the original document's name without the extension.
{n} — a running number for the part, starting at 1.
{page} — the first page of that part, the only reading of it that survives a chunked split.
{date} — the date, for exports you re-run on a schedule.

Bookmarks, metadata and what carries over

What carries into each output is everything page-level: the pages' content, fonts, images, dimensions, rotation, and the annotations and form fields that live on those pages. What does not carry is document-level structure. The bookmark outline is the visible one — its destinations point at page numbers that may not exist in a two-page extract, and an outline that looks intact but lands on the wrong pages is worse than none. The document's info header does not transfer either: each output is a fresh document, and the original Title and Author stay with the original.

Which is the other half of the honest answer: your original file is not modified at all. It is read, the parts are written as new files, and the document on your disk — bookmarks, metadata, everything — stays exactly as it was.

Extracted, not hidden

The failure mode worth knowing about in any splitting tool: an implementation that copies the whole document and merely hides the unselected pages produces a file that looks right and still contains every removed word, recoverable with any text extractor. It is the same class of mistake as redacting with a black rectangle, and it is how redaction failures make the news. Here the outputs are new documents with nothing to recover — and you do not have to take this page's word for it: run any text extractor over an output and search for a phrase from an unselected page. It will not be there.

Frequently asked questions

Is my PDF uploaded when I split it here?

No. The document is read, split and written back inside your browser tab by a Web Worker; there is no upload endpoint. Check it in DevTools: the Network panel stays silent while you split. The page's Content-Security-Policy sets connect-src 'self', so the browser itself would block any attempt to send the file out.

What does the range 12- mean?

Page 12 through the end of the document. Open ranges work both ways: -5 means the first five pages. Numbers are one-based, matching what the page selector prints, and the preview beside the input shows the resolved page list before you run the split.

Can I get the output pages in a custom order?

Yes — the output follows the order you wrote, so 5, 1, 3 produces a document in exactly that sequence, and duplicate page numbers are allowed. If rearranging is the whole job rather than a by-product of extracting, the reorder tool handles it more directly.

Will the split files keep their bookmarks?

No, and deliberately. A bookmark outline is document-level structure whose destinations point at specific pages; in an extract containing a subset of those pages, most destinations point at nothing. A broken outline that looks intact is worse than its absence. Annotations and form fields that live on the kept pages do survive.

Does splitting modify my original PDF?

No. The original is read only; each output is a new document downloaded alongside it. Your file keeps its bookmarks, its Title and Author metadata and everything else untouched — the parts are fresh documents, not edited copies of the original.

How many files can one split produce?

As many as the job needs, within the tool's 100 MB per-file input ceiling: Every page produces one file per document page, Chunks produces one per group of N, Ranges produces one. Memory is the real limit, since every output is built in the tab — the tool declines oversized work rather than freezing mid-split.

ToolsByUs team · How we check our tools