PDF Splitter

Split a PDF into individual pages or custom page ranges and extract the sections you need. All processing runs in your browser — no uploads required.

Advertisement

Split a PDF or pull out specific pages, without uploading the file

Drop a PDF into the box above and the tool reads its page count. Then choose one of two modes: split into individual pages, which produces one single–page file per page, or split by page ranges, where you type exactly which pages you want and get one file per range. Each result appears in a list with its size and its own download button.

The split runs entirely inside the browser tab. The file is read with the browser's file API, cut apart by a JavaScript PDF library that loads with the page, and the pieces are handed back as local downloads. There is no upload anywhere in the code — no request carries your document off the machine, so there is no server copy to be stored, indexed, or deleted afterwards. After the page loads you can disconnect from the network and it still works. When the thing you are cutting up is a client contract, a patient record, an ID document, or a payroll run, that is usually the deciding factor.

Page range syntax, precisely

Most of the value here is in the range box, so it is worth being exact about what it accepts. Enter a comma–separated list. Each item is either a single page number or two numbers joined by a hyphen. Every item produces its own output file.

You typeYou get
1-3, 4-6, 7-10Three files: pages 1–3, pages 4–6, pages 7–10
5One file containing page 5 only
1, 7, 12Three separate single–page files
1-2, 9Two files: a two–page file, then a one–page file
3-3One file with page 3 — same as typing 3
1 - 4Spaces around the hyphen are fine; one file, pages 1–4

Numbering is 1–based and matches what you see in your PDF viewer's page box — the first page is 1, not 0. Ranges are inclusive at both ends, so 4-6 is three pages.

The rules the parser enforces

  • Ranges must ascend. 9-3 is rejected. To reverse page order, list the pages individually — and note you would get separate files, not one reordered document.
  • Every page must exist. Any number above the document's page count is rejected, which catches the classic off–by–one of typing 1-25 for a 24–page file.
  • Validation is all–or–nothing. One bad item rejects the whole input; nothing is split, and the error message repeats your document's page count so you can correct the number without hunting for it.
  • Order follows your typing. 10-12, 1-2 produces the 10–12 file first. The list is not sorted for you.
  • Overlaps are allowed. 1-5, 3-8 is accepted, and page 3 through 5 appear in both outputs. Useful when two recipients each need an overlapping slice.
  • Gaps are allowed. Nothing requires you to account for every page. 2, 17 is a perfectly good instruction that ignores everything else.
  • No open–ended or shorthand forms. 5-, -5, last, odd, even, and all are not understood. To reach the end of the document, type the actual final page number — the tool shows it, and the placeholder in the box is pre–filled with your file's real page count.

Individual pages or ranges: which mode you want

The two modes solve different problems and the wrong choice is mostly a nuisance rather than a mistake.

Individual mode ignores the range box entirely and cuts every page into its own file, labelled Page 1, Page 2, and so on. It is the right answer when a downstream system wants one page per document — a records system that files each sheet separately, a batch of scanned forms that need to be routed to different people, or a stack of certificates where each recipient gets their own page.

The thing to know before you press it: a 200–page document produces 200 files, each with its own download button, and there is no ZIP and no download–all. That is a lot of clicking, and browsers frequently prompt or throttle when a page tries to save many files in quick succession. For anything beyond a couple of dozen pages, either use range mode to produce a handful of files, or accept that you will be working through the list.

Range mode is the better default for almost every human–driven task, because real tasks are about a specific slice, not about atomising the document.

Extracting one signed page from a contract

The most common request of all: the counterparty wants proof of execution, not the twenty–page agreement. Open the file, find the signature page in your viewer — say it is page 18 — select range mode, type 18, split, download. You get a one–page PDF containing exactly that page, and the twenty–page original is untouched on disk. If the execution block runs onto a second sheet with the notary stamp, type 18-19 and you get both in one file rather than two.

Two details matter here. First, check whether the page you want carries a needed exhibit reference on the sheet before or after it; a signature page in isolation sometimes loses the context that makes it meaningful. Second, if you want the page removed from the document rather than extracted, express it as the ranges you want to keep: to drop page 18 of a 24–page file, type 1-17, 19-24. That gives you two files, which you then combine — there is no delete–page operation here.

Separating a scanned batch

A flatbed or sheet feeder run produces one PDF containing several unrelated documents — three invoices, or a stack of signed consent forms, or a month of receipts. Scroll through once and note where each document starts, then type the boundaries in one go: 1-4, 5-9, 10-11, 12-18. Four files come back, in that order, each ready to file under its own name.

Scanned pages behave identically to text pages here — the split copies page content as it stands. If the scan carried an OCR text layer, the pieces keep it and remain searchable; if it did not, they do not, because nothing in this tool performs OCR. Pages also keep their original size and rotation, so a sheet that fed in sideways comes out sideways in its extracted file. Fix rotation in the source before splitting if it matters.

Other tasks this handles cleanly

  • Sending one chapter or section of a long report to someone who should not see the rest.
  • Getting under an upload limit by submitting a portal's required pages instead of the whole file.
  • Splitting a combined bank or card statement into per–month files for an accountant.
  • Pulling the pages a redaction review actually needs, so a reviewer is not handed the entire document.
  • Separating a two–sided scan where fronts and backs landed in one file, by listing the pages you need in the order you need them — accepting one file per page.

What the extracted pages carry with them

Each output is a newly built PDF containing copies of the pages you selected. Page content transfers faithfully; document–level structure does not, because it belonged to the original document as a whole.

ElementIn the extracted file
Text, vector graphics, imagesPreserved exactly — nothing is re–rendered or re–compressed
Embedded fontsPreserved, so the page renders correctly elsewhere
Page size and rotationPreserved per page
OCR text layer on scansPreserved if it was there
Bookmarks / outlineNot carried over
Interactive form fieldsDo not survive as working fields — flatten the form first if the values matter
Document metadataNot carried over
Links pointing to pages outside the extracted rangeBroken, necessarily — the target is not in the file

Page numbers printed on the page do not change. If you extract pages 18–19, the resulting two–page file shows "18" and "19" in the footer, because that text is part of the page content. This is usually what you want for a legal or audit context — the extract is evidently a slice of a larger document — but it surprises people who expected renumbering. Nothing here renumbers pages.

Limits and errors

  • PDFs only. A file the browser does not report as application/pdf is refused with a message.
  • Encrypted and password–protected files cannot be opened, and you will see a message saying the file may be corrupted or encrypted. Save an unprotected copy from a viewer that accepts the password, then split. Permissions–only passwords count too.
  • The split button stays disabled in range mode until you type something.
  • One source file at a time. Use Reset to start over; it releases the previous results, so download what you need before resetting.
  • Downloads are named after the source file plus the page or range, so several extracts from one document do not collide in your downloads folder.
  • No ZIP archive. Every output is downloaded individually.
  • Memory is the real size ceiling. The source and every result are held in the tab at once, which matters most in individual mode on a large scanned document — consider doing it in two range–mode passes instead.
  • The original is never modified. The file on your disk is read, not written; every output is a new document.

Can I reorder pages while splitting?

Not within one file. Each range produces its own document with its pages in ascending order. Reordering means extracting the pieces here and then combining them in the order you want, which is a separate operation.

Does splitting reduce quality or file size?

Quality, no — content streams are copied verbatim. Size is less predictable than people expect: a single extracted page is usually far smaller than the original, but it is not the original divided by the page count. Resources the page depends on, most notably embedded fonts, must be carried into every file that needs them, so a one–page extract from a font–heavy report can be larger than a simple arithmetic share suggests. Extracted scan pages track the size of their own bitmap closely.

Why did my range get rejected?

Almost always one of four things: a number larger than the document has pages, a descending range like 12-4, a stray character in the list, or a hyphen that is actually an en dash pasted from a document — type the hyphen rather than pasting the range from formatted text. The error message includes your file's page count, which resolves the first case immediately.

This tool is provided for informational and educational purposes only. All processing happens in your browser — no data is sent to or stored on our servers. While we strive for accuracy, we make no warranties about the completeness or reliability of results.
PDF Splitter | InventiveHQ