Split PDF by Range, Bookmarks, Text or Blank Pages

About Split PDF by Range, Bookmarks, Text or Blank Pages

Splitting a PDF comes down to deciding where the cuts go. Besides the plain modes — every N pages, a page range, or a size limit — this tool gathers the smarter ones in one place: cut at bookmarks, at a value printed on the page (an invoice number, an employee ID), at blank sheets or QR separator sheets in a batch scan, and a mode that splits two-page book spreads down the middle.

Every mode builds its output by moving page objects across, never by re-rendering. Text stays text, vectors stay vectors, and nothing loses quality. That approach has one visible side effect: the split files together weigh more than the original.

Why the pieces add up to more than the whole

Fonts, logos and colour profiles are stored once in a PDF and shared by object number across every page that uses them. If a 100-page report shares a single 1.5 MB embedded font subset, then cutting it into ten files means each file needs its own full copy of that font. And the subset holds every glyph used anywhere in the original document, so a three-page extract still carries the characters for all 100 pages.

For each output, the tool deletes objects its pages no longer reference — content of pages left out, images only those pages used — before saving. Only the shared resources repeat; each page’s own content appears once. A scanned PDF, where every page is its own image, splits into files that total roughly the original size; the heavier the fonts, the bigger the overhead.

How the size-limit mode hits its target

“No file over 10 MB” can’t be turned into a page count: the first three pages might be a photo-heavy cover and the rest plain text. This mode adds up the size of the objects each page references to estimate how many pages fit, then corrects the next estimate by how far the last saved file was from its prediction. If a saved group still comes out over the limit, only that group is binary-searched down to the largest page count that fits.

A single page that exceeds the limit on its own — an 8 MB scan, say — can’t be split, so the tool stops and names the page; compress the images first. For email, aim 5–10% under the provider’s cap: attachments are Base64-encoded in transit and grow by about a third, which is why a 25 MB limit rarely accepts a 25 MB file.

Splitting at bookmarks

Bookmarks form a tree of titles, each pointing at a page. If level 1 holds chapters and level 2 sections, choosing level 1 gives one file per chapter, named after the bookmark (“001-Introduction.pdf”). Only bookmarks that actually point at a page count as split points; title-only grouping entries and bookmarks that open a web address are skipped.

Each output keeps just the bookmarks inside its own range, and entries pointing at pages that are no longer there are removed. A parent heading with children stays as a heading even if its own target page left.

Splitting by text: batched invoices and payslips

A 300-page run of invoices exported from accounting software usually repeats a line like “Invoice #: INV-1024” on every page. This mode reads each page’s text layer and cuts wherever that value changes, or wherever it appears. When you open a file, it looks for “label: value” lines that recur on many pages and change often, and offers them as suggestions; something identical on every page, like your company name, can’t split anything and ranks low.

Restricting the search to the top N percent of the page stops a reference number quoted in the body from triggering a cut. File names accept tokens such as {value}, {n} and {start}, so the output can be named “INV-1024.pdf” and sent as is. A scanned PDF has no text layer; run OCR first.

Splitting at blank pages or QR separator sheets

Feed a stack of separate documents through a sheet-fed scanner with a blank sheet between them, and this mode cuts at the blanks. A page counts as blank when the share of dark pixels — brightness below 160 on a 0–255 scale — is at or below 0.05%, measured with a 5% border ignored on every side. The border is excluded because edge shadows and punch holes are what usually turn a blank sheet into a “page with content”.

In a duplex scan one separator sheet produces two blank pages, so set the minimum run to 2; a single blank back side inside a real document then won’t be mistaken for a separator. Printed QR separator sheets are more reliable still, and their text (for example “DOC:Acme-contract”) can become the file name. Separator pages can be dropped, or kept as the first page of the next document or the last page of the previous one.

Splitting book spreads

Scanning an open book puts the left and right pages on one wide sheet. This mode doesn’t cut pixels. It makes two shallow copies of the page dictionary that share the same content stream and narrows each one’s CropBox — the visible area — to the left or right half. Quality is untouched and the file barely grows, but the hidden half of each image is still in the file.

The split line can move between 20% and 80%, and each half can overlap the line by a few millimetres so text near the gutter isn’t clipped. For right-to-left books (Japanese vertical text, Arabic, Hebrew) choose right-first, and leave the first and last sheets whole if the covers were scanned as single pages.

Worked examples

How to use it

1. Choose a PDF and a split method: pages/range/size, bookmarks, text, blank pages/QR, or spreads.
2. Set the criterion and check the proposed cuts in the preview; toggle individual pages on or off as boundaries if needed.
3. Run the split and download the files one by one or as a ZIP.

Frequently asked questions

Related guides

• How PDF Files Work: Objects, the Cross-Reference Table, and Why PDFs Are Hard to Edit