Split PDF

The distinction that matters in this mode is between a list of pages and a list of groups. Typing 1-3, 5, 7-10 does not select nine pages — it describes three documents, and three PDFs come back: one holding pages 1 to 3, one holding page 5, one holding pages 7 to 10. Every comma is a cut.

That makes ranges the right mode whenever the document already has a structure you can name. A report whose chapters begin at pages 1, 12, 29 and 44 splits into four chapter files from a single line of input. The alternative — extract one section, save it, reload the original, extract the next — is the same work done four times over, and it is where the page numbers usually get transposed.

The merge option changes the result rather than the selection: the same groups are collected into one PDF instead of several. That is how you drop a section out of the middle of a document — name the ranges either side of it and turn merging on. Either way pages come out in document order rather than the order you typed them, a page named twice appears once, and anything producing more than one file arrives as a ZIP with the page numbers in each filename.

This mode takes one number and does the arithmetic for you. Set 15 pages per file and a 553-page document becomes 1-15, 16-30, 31-45 and so on up to 541-553 — thirty-seven files, no ranges typed. The last part holds whatever is left over, so it is usually shorter than the others, and the file count is shown before you commit to it.

It is the right mode when a document has no structure worth naming. Scanned stacks, exported logs, bulk downloads and archive files rarely have chapters; they have pages, and the only sensible cut is a regular one. It also suits a constraint expressed in pages rather than megabytes — a portal that accepts twenty pages per submission, or a print job that has to go in batches.

Setting it to 1 saves every page of the PDF as its own document, delivered together in a ZIP. That is the answer to a batch scan in which each sheet is really a separate artefact — invoices, certificates, delivery notes, signed forms — fed through in one pass because doing them individually would have taken an afternoon. When you want only some of the pages separated rather than all of them, Extract Pages does the same thing to a selection.

Page count is a poor proxy for file size. A page of body text might be four kilobytes and the scan on the page after it four megabytes, so 'ten pages per file' tells you nothing about whether the result will clear a 25 MB mail server. Splitting on size inverts the question: you supply the cap, and the tool works out how many pages fit under it.

Working that out honestly is the whole trick. The obvious method — measure each page, add the sizes up until you reach the limit — is wrong, because the fonts, colour profiles and images in a PDF are shared between every page that uses them. Saved alone, twenty pages each carry their own copy of the same font; saved together, they carry one. Adding individual sizes therefore counts the shared parts over and over: on a 35 MB document of 553 pages it over-counted by roughly four times and produced thirty tiny files where about eight were right.

So each candidate group is genuinely built and measured rather than estimated, grown by doubling and then narrowed by bisection, which costs a handful of saves per part instead of one per page. That work runs on our servers rather than in the tab, because hundreds of re-saves would freeze the browser. Parts are packed to roughly 94% of your cap — a re-save varies by a little and 'MB' means 1,000 KB in some tools and 1,024 in others, so a part reporting 5,073 KB against a 5 MB limit would be technically correct and practically useless. A single page larger than the cap cannot be made smaller by cutting around it and is emitted whole in its own file; switching on compression runs a light, text-preserving pass over the document first, which is usually enough to bring scanned pages under the line.

The page picker shows every page as a thumbnail, and clicking one selects it. What you click is mirrored into a range box as you go — pick pages 4, 5, 6 and 9 and it reads 4-6, 9 — so the selection can be checked as a line of text instead of by scanning a grid for highlighted tiles. In a long document that reading is what catches the page you meant to click and missed.

Two things can then happen to that selection. By default it becomes one PDF holding those pages in document order, which is what you want when the pages belong together: a section, a run of related forms, the three pages somebody actually asked for. Switch on 'Save each page as a separate PDF' and each selected page becomes a file of its own inside a ZIP, which is the shape you want when every page is a document in its own right.

This overlaps with two neighbouring tools on purpose, and the right one is whichever leaves you doing less. Extract Pages is the same operation with the whole screen given over to choosing pages, so it is the better place to start when picking is the entire job. Delete Pages is the complement, and it is quicker whenever naming what goes is easier than naming what stays — which it is whenever you are keeping most of the document. Staying here makes sense when you are already splitting and one section needs pulling out as well.