Split by Bookmark

Split by Bookmark reads the document's outline with qpdf, takes the bookmarks sitting at the level you asked for, and turns each one into a PDF of its own. The page a bookmark points at becomes the first page of its part, and that part runs to the page before the next bookmark begins; the final one runs to the end of the document. Each part is written by copying its page range out of the original rather than re-rendering it, so text stays selectable and a scan keeps the resolution it arrived at.

Bookmark level to split at decides how finely the document is cut. Level 1 uses the top row of the outline, which is usually the chapters, and level 2 goes a row deeper into the sections beneath them, as far as level 6. If the level you pick holds no entries, the job falls back to the top row rather than failing, so an over-deep guess still returns something. Name each file after its bookmark is on by default and puts the outline title into the filename behind a two-digit index, as in 01 Introduction.pdf; switched off, the parts are numbered after the source file instead.

The limits are all about where a cut can land. A part begins at a whole page, so a chapter that starts halfway down a page brings the tail of the previous one with it, and anything above the first bookmark at the level you chose — a cover, a contents list, or at level 2 the pages a chapter opens with before its first section — belongs to no part at all. Boundaries one row up are not boundaries here either: cutting at level 2 runs the closing section of one chapter straight on into the next chapter's title page, because only the level you asked for is read. Two bookmarks resolving to the same page leave the earlier of the pair with an empty range, so it writes no file. Titles are reduced to ASCII letters, digits, underscores, spaces and hyphens for the filename and cut at 60 characters, so punctuation vanishes, an accented letter goes with its accent — Café becomes Caf — and a heading with nothing left after that filter is written as section- plus its number.

Processing is done by the OmniKitly service, not in the page. Your PDF is uploaded over an encrypted connection to a machine running qpdf, and the upload, the parts it produced and the job record are removed together once the retention window passes — hours, not days. Three ceilings apply before that: 200 MB per upload, 300 pages per document and 90 seconds of processing. A PDF carrying no outline is refused with a message saying exactly that rather than handed back whole, and one that demands a password cannot be opened to read its bookmarks at all.

Secure server processing. This operation requires temporary server processing.