Split PDF pages
One document in, the pieces you asked for out. Runs in your browser — no upload, no account.
Runs in your browser0 bytes uploadedDrop your PDF here
Nothing is uploaded. The pieces are cut here, from bytes that never leave this tab.
How to split PDF pages without handing over the whole file
There is an awkward shape to the usual advice here, and it is worth naming before anything else. The reason most people split a document is that only part of it should travel: two pages of a medical record for an insurer, the signature page of a contract, one invoice out of a scanned stack of forty. Every mainstream tool for that job begins by asking you to upload the entire document to a server — including the pages you were specifically trying not to send. The extraction happens after the disclosure it was meant to prevent.
Here the order is reversed, because the document never has to be anywhere except where it already is. A PDF is not a picture but a numbered tree of objects, so the library running inside this tab walks that tree straight out of the bytes your browser read off your disk, picks up the sheets you named, and assembles fresh documents around them. That is bookkeeping, not rendering. It finishes instantly on hardware a decade old, and routing it through a data center would buy you a queue, a wait, and somebody else's storage holding your paperwork.
Any site can type the words in your browser, so treat ours with suspicion and go and look. Open the developer tools, switch to the network tab, hand in a document and cut it while watching: the page pulls down its own code once, and then the tab falls quiet. What you download afterwards was assembled from memory belonging to this tab alone. Hand in one document — that first act is what fetches the PDF library — and then disconnect from the network completely: everything after that still works, because the only thing this address ever requests is its own code.Auditing that kind of claim on any site, ours very much included, is a short exercise with a page to itself.
PDF, in brief
| Full name | Portable Document Format |
|---|---|
| Extension | .pdf |
| Format type | Fixed-layout document |
| MIME type | application/pdf |
Four ways to cut, and why there are not five
Splitting sounds like one operation and is really four different questions, so the tool asks which one you have rather than pretending a single control covers them all.Split in two is for a document with one boundary you can point at — the report ends and the appendix begins. Every N pages is for material that is secretly a batch: a scan of ten two-page invoices is one file that should have been ten. One file per page is for pages heading to different destinations. And custom ranges takes what you type, where each group becomes its own file: 1-3, 7-9 gives you two documents, not one.
There is deliberately no fifth mode for extracting pages, because extraction is custom ranges with the put them all in one file box ticked. Two controls that perform one operation is how a tool starts to feel arbitrary. For the same reason the slider for a cut point stops one page short of the end and the slider for chunk size stops one page short of the whole document: at those values the operation hands you back the file you started with, and an option that does nothing should not be on screen.
Both sliders have a number box next to them. That is not decoration either — a slider alone makes page 47 of a 300-page document essentially unreachable, and a default is supposed to be a shortcut rather than a ceiling. The range field takes the shapes people really type rather than one rigid grammar: 1-3 for a span, 7- to run from page seven to the end, -4 to stop at four, last for the final sheet without counting to work out which one it is, and the en dash that word processors quietly substitute while you type.
Before anything runs you see the list of files that will be produced, each with its name and its exact page count. If your ranges leave pages out — 1-3, 7-9 in a nine-page document abandons four, five and six — the tool says so in advance. That is the most dangerous failure in this category precisely because it does not look like one: you get files, they open, they are correct, and the missing pages are noticed weeks later by somebody else.
Why the pieces add up to more than the file you started with
This surprises people, and the honest answer is more interesting than an apology. Pages in a PDF share things: an embedded font, a logo drawn on every sheet, a color profile. A page cannot reference a font that lives in a document it is no longer part of, so each piece has to carry its own copy of whatever its pages relied on. Measured on our own test files, cutting one document into one file per page produced totals of 2.2 and 2.5 times the original. Splitting into a few large pieces costs very little; splitting into many small ones costs the shared material over and over.
The same arithmetic is why you will not find a size estimate above the button. Weight inside a PDF is not spread evenly across its pages, and we measured how badly: in a four-page test file of 302 KB, page two came to 99.8 percent of the total because it holds a scan, while each of the other three was half of one percent. A forecast that divided the file by its page count would have said 25 percent for every page and been wrong by a factor of two hundred on three of the four. A dishonest number is dishonest even when it is displayed in a tidy monospace column, so what appears in advance is what is genuinely knowable — which pages land in which file — and the real sizes appear afterwards, measured.
The two PDF passwords, and why only one should stop you
A protected PDF is not one thing but two, and the difference decides whether you are inconvenienced for a good reason or for no reason. An owner passwordonly asks readers not to print or copy; the file opens for anybody, and the restriction is a request that software chooses to honor. A user password is real encryption, and without it there is no way in and we do not pretend to build one.
The two are indistinguishable from the outside — we measured a matched pair of files and both reported identical encryption dictionaries — so the only honest test is to try the door with an empty password. A file carrying only an owner password opens; a file with a user password answers that it needs one. Doing that costs a few milliseconds and means you are never asked for a password you do not have and do not need. When a file does need one, the box appears and nothing is split until the password actually opens the document.
What will never happen here is the shortcut most PDF libraries advertise: a setting that suppresses the encryption test while every content stream stays scrambled. Documents built that way open cheerfully, count their pages correctly, and hold nothing anyone can read, because the ciphertext was copied straight through into a container that claims to be plain. We ran that experiment before this page was written. If you would rather deal with a lock once instead of on every operation, removing the password from a PDF is its own page, and it is also where you find out which of the two you are facing.
What this page does not do
It only cuts. Putting several documents together runs the other way and has an address of its own instead of a switch on this one, so that anybody following a link to a splitter arrives at a splitter: merge PDF files handles that direction. Both get reached for in the same sitting surprisingly often — lifting pages 3 to 8 out of one report and 1 to 4 out of another, then joining the results, is this tool followed by that one.
Nothing is redrawn on the way through. Dimensions, typefaces and artwork reach the pieces exactly as they left the source, with no second round of compression, so a sheet that was A4 inside a document mixing A4 with Letter is still A4 once it stands alone. Words you could select and search stay selectable and searchable, which is the property that disappears the moment a tool rebuilds pages by photographing them — and its absence goes unnoticed until somebody tries to quote a paragraph. PDF to textis a fast way to audit what came through, whilehow to copy text from a PDF covers why some documents carry a text layer and others never had one. When you want pictures rather than documents out of the sheets, PDF to PNG renders them.
One casualty is real and deserves a plain sentence: a fillable form stops being fillable. Fields do not belong to the sheets — they hang off the document as a separate index — and every piece is a new document that index does not travel to. Whatever somebody had already typed remains visible, because it was drawn onto the page as well, so the answers persist as ink even though nobody can edit them afterwards and no system reading submissions by field name will find any. A count of those fields appears before you run anything. Bookmarks and the outline vanish for the same structural reason and, unlike the answers, leave nothing behind. No character recognition happens here either, so pieces taken from an image-only scan remain image-only.
What will never happen is a refusal. Neither length nor weight gets anybody turned away: hand over something enormous and you get a plain note about the memory the operation is likely to want, and then the operation. The machine belongs to you, and so does the judgment about what it can take.
Questions
Does my PDF get uploaded?
No. The file is read off your own disk, the pieces are assembled in memory, and the links you download from point at data that exists nowhere but this tab. You can confirm that rather than believe it: bring up the network panel, run a split, and confirm that not one request carries a page of it. It matters because of what people split — contracts with an appendix that belongs to somebody else, a scanned stack of invoices, a medical record where only two pages should ever be forwarded. That last case is the one worth pausing on, since uploading the whole file in order to extract two pages hands over exactly the pages you were trying not to share.
How do I split a PDF into multiple files?
Four ways, and each is a different question. Split in two after a page, when there is one boundary you can point at. Every N pages, when a scanned batch is really ten two-page invoices. One file per page, when the pages are going to different places. Or type your own ranges as 1-3, 7-9, where each group you type becomes its own file. The list of files that will be produced appears before you run anything, with the exact page numbers in each, so nothing about the result is a surprise.
Can I extract pages instead of cutting the file up?
Yes — type the ranges you want and tick "Put them all in one file instead". You get a single PDF containing pages 1-3 and 7-9 and nothing else, rather than two separate documents. That is the same operation as extracting pages, which is why there is no separate mode for it: two buttons that do one thing is how a tool starts feeling arbitrary.
Why is there no size estimate before I run it?
Because any number we could show you before doing the work would be a guess dressed up as arithmetic. Size in a PDF is not spread evenly across pages. We measured a four-page test file of 302 KB: page two carried 99.8 percent of it because it holds a scan, and each of the other three came to half of one percent. A proportional forecast would have said 25 percent for every page and been wrong by a factor of 200 on three of the four. What is genuinely known in advance is which pages land in which file, and that is shown exactly. Real sizes appear on the receipt, measured rather than predicted.
Why is the total larger than the file I started with?
Because each piece has to be able to stand on its own. Pages share things — an embedded font, a logo, a color profile — and one page cannot reference a font that lives in a document it is no longer part of, so every output file carries its own copy. Measured on our own test files, one file per page came to 2.2 and 2.5 times the original. Splitting into a few larger pieces costs far less than splitting into many small ones, and no arrangement makes the total smaller than the source.
What happens to a password-protected PDF?
It depends which of the two passwords the file carries, and most tools do not distinguish them. An owner password only asks readers not to print or copy; the file opens without anyone typing anything, so it is split without asking you for a password you may not have. A user password is real encryption, and there the file gets a password box and nothing is split until the password actually opens it. What never happens is the shortcut most libraries advertise: a setting that suppresses the encryption test while leaving every content stream scrambled. It yields documents with correct page counts and nothing anyone can read, and it fails without saying so.
Is there a page or file size limit?
None is imposed. A large file gets a note about the memory the work is likely to need, and then the work runs, because the hardware is yours and so is the decision. A 900-page document is not an error. Memory rather than time is the real wall, the original, every piece and the preview occupy the tab simultaneously, so one file per page on a very long document is the arrangement most likely to make it struggle, and splitting it in two passes costs nothing.
Does splitting change how the pages look?
No. Dimensions, typefaces and artwork reach the pieces exactly as they left the source, with no second round of compression, so a sheet that was A4 inside a document mixing A4 with Letter is still A4 once it stands alone. Words you could select and search stay selectable and searchable, which is what disappears the moment a tool rebuilds pages by photographing them. One thing genuinely does not survive: fillable fields stop being fillable, since they hang off the document as a separate index rather than belonging to the sheets. Whatever somebody typed in remains visible because it was drawn onto the page as well — the answers persist as ink, the form stops being a form.
How is this different from merging PDFs?
Merging runs the other way and has an address of its own instead of a switch on this one, so that anybody following a link to a splitter arrives at a splitter. Both get reached for in the same sitting surprisingly often: lifting pages 3 to 8 out of one report and 1 to 4 out of another, then joining the results, is this page followed by that one.