Split PDF

The common case is one file per page: a scan of a dozen receipts that a form wants submitted separately, or a bundle of statements that need filing individually. Leave the range box empty and that is what you get, named with a padded number so they sort correctly in a file manager rather than putting page 10 between 1 and 2.

The other case is a long document that is really several — a report whose three chapters go to three people. Naming ranges gives you one file per range instead, with the pages in document order.

The pages are copied rather than redrawn, so each piece is your document rather than a picture of it. Text stays selectable and searchable, vector artwork stays sharp at any zoom, and images come across byte for byte. There is no quality setting because there is no quality decision to make.

What does not survive is the same as with merging, and for the same reason: bookmarks, form fields and the document outline point at pages by reference, and splitting renumbers everything underneath them. If your document depends on those, use a full PDF editor.

Splitting takes one document apart and gives you all of it back, in pieces. Extracting takes a few pages out and gives you only those, leaving the original alone. If what you want is pages 4 to 6 of a forty-page report and nothing else, the extract page is the one that does it in a single step; doing it here means splitting into forty files and throwing away thirty-seven.

The distinction matters because the two read almost the same in a search box and produce very different amounts of work. This page is for when the whole document needs to become several documents.

Each file carries the source name and the pages it holds, with the number padded — `report-003.pdf` rather than `report-3.pdf`. That padding is not decoration: a file manager sorts names as text, so without it page 10 lands between page 1 and page 2, and a folder of thirty pieces comes back in an order nobody wants.

When you name ranges, the pieces are numbered by range rather than by page, in the order you wrote them. The names are the only record of what came from where, since a split piece has nothing inside it that says which document it used to be part of.

Ranges are page positions, not content. `1-3, 7, 10-12` works, `8-` means page eight to the end, and a backwards range like `7-3` is read as the same pages rather than refused. A page number that does not exist stops the job with a message instead of being quietly skipped, because a silently dropped page is the failure you would find out about later.

What a range cannot do is find anything. "Every page with an invoice number on it" or "each chapter" are content questions, and this tool never looks at what a page says — only at where it sits. If the split points depend on the text, you need to read the document and write the ranges yourself.

A few hundred pages is unremarkable. The work is proportional to the number of pieces rather than to the length of the document, so splitting a 400-page file into 400 files is the demanding case and splitting it into four is not.

The limit is your own memory, because every piece is assembled before the ZIP is written. On an older machine, a very large scan split page by page is the combination to watch — do it in ranges of fifty and save as you go. There is no server-side cap to run into, because there is no server.