Extract Exactly the Pages You Need
Most documents that cross an office desk are bigger than the part that matters. A forty-page agreement where only the signature page needs forwarding; a scanned bundle with three blank pages the feeder picked up; a two-hundred-slide lecture deck of which one chapter is examinable; an exhibit that must be filed separately from the contract it was attached to. Splitting solves all of these with one input: a page range. The syntax is the one you already know from print dialogs — numbers and hyphenated spans separated by commas, so 1-3, 5, 8 selects pages one through three, five, and eight. After you drop a file the tool reports its true page count and validates your range against it, so an out-of-bounds request fails with a clear message instead of producing a wrong document.
The same range drives two complementary modes. Keep only these pages builds a new PDF from your selection — the classic extract. Remove these pages inverts the selection and keeps everything else, which is the natural way to drop a cover sheet, delete blank scans, or strip an appendix. A guard rail refuses any operation that would leave zero pages, so you cannot accidentally empty a document.
How Lossless Page Extraction Works
The naive ways to pull pages out of a PDF — printing to a new PDF, or screenshotting — both pass the
content through a renderer and back, turning crisp text into pixels. This tool does something
fundamentally different. The document is read into memory as an
ArrayBuffer and parsed by
pdf-lib into its native object graph.
The pages you selected are then copied verbatim — each page dictionary together with the fonts,
images and annotations it references — into a fresh document. Nothing is rasterized and nothing is
recompressed: text remains selectable and searchable, vector line-work stays sharp at any zoom
level, and embedded font programs survive byte-for-byte.
A pleasant side effect of object-level copying is that extracts are usually far smaller than the source. When the new file is serialized, only resources actually referenced by the surviving pages are written out; the high-resolution images living on the pages you left behind simply never make the trip. Pulling five pages from a 60 MB scanned binder routinely yields an attachment of a couple of megabytes — with zero quality compromise, because nothing that remains was altered.
Everyday Splitting Recipes
A few patterns cover most office needs. To forward just a signature page from a long agreement, extract that single page number. To drop a fax cover sheet, switch to remove mode and enter 1. To break a large report into chapters for separate reviewers, run the tool once per chapter range — each pass reads the original from your disk, so there is no degradation from repeated runs. To clean a scan, remove the blank page numbers the feeder inserted, then run the result through Compress PDF if it still needs to fit an email limit. And if the trimmed document needs continuous numbering, finish with Add Page Numbers — printed numerals are page content and never renumber themselves.