In short
This extracts pages from a PDF — a single range, or a full split into separate documents — entirely in your browser. It's built onpdf-lib, a pure-JavaScript library, so there's no server and nothing to install. The document is parsed locally and the pages you choose are copied into a new PDF you download. Nothing is uploaded, and the Network tab proves it.
How the split works
Splitting is page copying, not re-rendering. pdf-lib reads the source document's page objects and copies the ones you selected — together with the fonts, images and resources each page references — into a fresh document. Because it moves the actual page content rather than rasterising it, text remains selectable and searchable, images keep their resolution, and page sizes stay exactly as they were. Extract pages 3 to 8 and you get a clean eight-minus-two-page PDF; split every page and you get one file per page.
Why no-upload wins for splitting
You often split a document precisely because it holds more than you want to share — a single statement pulled from a year's worth, one signed page from a long contract, a section of a medical record. Uploading the whole file to a server to extract a fragment defeats the purpose: now the entire document sits on infrastructure you can't inspect. In-browser splitting keeps everything local — the file is read from disk, the pages are copied in your tab, and no copy leaves your machine. Confirm it yourself in the Network tab.
Encrypted inputs: unlock first
An encrypted PDF's pages can't be copied until the document is decrypted, so the tool won't quietly hand you an empty or broken split. It detects encryption and tells you to unlock the file first. Run it throughUnlock PDF — which lifts owner-password restrictions and decrypts empty- or known-password files locally — then split the unlocked copy here.
If the file is damaged, repair it first
Split can only extract pages it can read. If the PDF is structurally damaged — truncated, interrupted mid-save, or clipped in transit — pdf-lib won't parse it, and no honest tool can extract pages that can't be read. When the tool can't parse a file, it stops and points you to repair. Looks damaged? Diagnose & repair it, then come back and split the repaired copy.