Pages in—
Files out—
Largest—
Total out—
Most splitters cut where you tell them and hand you document_1.pdf through document_9.pdf. Two things are missing from that, and both are the reason people end up doing it by hand: a document that has chapters already knows where it should be cut, and a document being split to fit an attachment limit has to be weighed, not estimated. This one cuts at the bookmarks — naming each part after the heading it starts at — and reports the true byte size of every part, because it has already built them. Splitting by file size is a paid feature at iLovePDF; here it is a slider.
Step by step
01Open the document
Drop it anywhere on the page. Every page is drawn as a thumbnail, so you can see what you are cutting rather than counting page numbers in another window.
02Choose how to cut it
At bookmarks for a chaptered document, by page range for a specific extract, every N pages for a scanned batch, or under a file size for an attachment limit.
03Check the parts
Each part lists its page count, its real size, and how many bookmarks travelled with it. Click a page in the grid to move a cut; the ranges expression rewrites itself so the link still reproduces it.
04Take one part or all of them
Each row has its own download; the header button gives you the whole set as a zip. Nothing was uploaded, so both are instant.
A worked split
The intuition is that a 100-page, 20 MB document splits into four 25-page files of 5 MB each. It almost never does, and the reason is worth knowing before you promise someone an attachment that fits.
Given
Where the weight actually is
Embedded fonts ~400 KB shared by every page that uses them
Scanned images 18 MB concentrated on pages 12-31
Text and vectors ~900 KB spread evenly
Structure ~50 KB catalog, page tree, outline
A 25-page cut over pages 12-31 ≈ 14 MB ← nowhere near 5
A 25-page cut over pages 60-84 ≈ 300 KBA PDF's size is not distributed by page, so dividing the total by the page count produces a number that is wrong in both directions — some parts blow the limit, others waste it. There is also no formula for it: a font embedded once is carried by whichever part contains a page that uses it, so a part's size depends on which pages are in it. The only correct answer is to build the file and look, which is trivially cheap when the split is running on your own machine and expensive when it is running on someone's server — which is why the services that offer it charge for it. This tool grows each part by doubling and then bisecting, so it finds the largest page count that fits in about a dozen probes rather than a hundred.
Before you cut
A chaptered document already knows where its cuts go
The bookmark tree is a table of contents with page numbers attached. Cutting at its top level gives one file per chapter with no counting, and cutting a level deeper gives one per section. The parts can also be named properly, because each one starts at an entry that has a title.
Each part should keep its own bookmarks
A chapter extracted from a 400-page manual still has sections, and they are still worth navigating. The entries whose pages landed inside a part are renumbered and written into it, so a part opens with its own outline rather than an empty pane.
Size cannot be predicted, only measured
Fonts, images, and colour profiles are shared between pages, so a part's size depends on which pages are in it rather than how many. Any tool that reports an estimated size before building the file is guessing, and the guess is worst exactly where it matters — a document whose weight sits in a handful of scanned pages.
One term, one file
Typing 1-3, 5 means two files rather than one file of four pages. It is the difference between splitting and extracting, and treating the comma as a separator between outputs is what makes a single expression able to describe a whole batch.
Splitting never re-renders the pages
Page content, fonts, and images are copied across byte for byte, so a part is visually identical to the pages it came from. Nothing is rasterised, nothing is recompressed, and text stays selectable and searchable.
The free tiers elsewhere are metered, and this is why
Sejda allows three tasks an hour and caps free documents at 200 pages or 50 MB; iLovePDF marks size-based splitting Premium. Those limits pay for server CPU and storage. A split that runs in your browser has neither cost, so it has neither limit.
The judgement call
What you are trying to end up with, and the mode that gets there.
A manual or report with chapters
At bookmarks
The cuts are already in the document, and each part comes out named after its own heading rather than numbered.
Sending one section to someone
Page ranges
A single term like 12-19 produces exactly one file and leaves the rest of the document out of it.
A scanned batch of identical forms
Every N pages
Two-page forms scanned end to end split cleanly on a fixed stride, with the short last part reported rather than hidden.
Attachments that must clear a mail limit
File size
It builds and weighs each part, so the sizes in the list are the sizes that will arrive rather than a division of the total.
A document with no bookmarks
Not at bookmarks
There is nothing to cut at, and the tool says so rather than silently handing back the whole document as one part.
One page that is bigger than your size limit on its own
Compress it first
No split can make a single page smaller. A part that is over the budget alone is reported as such rather than being dropped.
Reference
FAQ
Open the document in a splitter that reads the outline, choose the bookmark level to cut at, and take the parts. The Tinapps split PDF tool cuts at the outline's top level by default — one file per chapter — names each part after the heading it starts at, and carries that chapter's own sub-entries into the file so it opens with its sections still listed.
Yes, but only by building the files and weighing them. A PDF's size does not divide by its page count, because fonts and images are shared between pages, so the size of a part depends on which pages are in it. The Tinapps split PDF tool grows each part until one more page would exceed the budget and reports the real byte size of every one — a mode iLovePDF charges for, because on a server it costs real CPU.
No. Splitting copies page content streams, fonts, and images across unchanged, so the pages in a part look exactly like the pages they came from and their text stays selectable and searchable. Nothing is rasterised or recompressed, which is also why the parts together usually weigh slightly more than the original — shared resources are now stored once per part.
In the Tinapps split PDF tool there is no page limit and no task counter, because the split runs on your own machine rather than a server. For comparison, Sejda's free tier caps a document at 200 pages or 50 MB and allows three tasks an hour, which is what paying for server time looks like.
In most tools they are discarded entirely, so each part opens with an empty navigation pane. The Tinapps split PDF tool keeps the entries whose pages landed inside a part and renumbers them to their new positions, so a chapter extracted from a manual still lists its own sections.
Type the pages you want as a single range term. In the Tinapps split PDF tool an expression like 12-19 produces one file containing those eight pages and nothing else, and the tool reports how many pages of the original were left out so an accidental omission is visible.
With most services, yes — the document is sent to a server, processed there, and deleted on a schedule you have to trust. The Tinapps split PDF tool never sends it: pdf-lib splits the file inside the page and pdf.js draws the thumbnails, so the document stays on your device and the tool keeps working with the network disconnected.
Because a single page can weigh more than the budget on its own — a full-page scan at 600 dpi easily exceeds 5 MB — and no split can make one page smaller. The Tinapps split PDF tool reports those parts instead of dropping the page or silently exceeding the limit; compressing the document first is the fix.
The document is read, cut, and weighed entirely in your browser. Nothing is uploaded, no part is stored anywhere, and the tool keeps working offline once the page has loaded.
Keep going
Merge PDF
The other direction, with the same bookmark handling and a report of what it costs.
PDF tools
The rest of the category — rotate, extract, delete, and convert.
Image compressor
Shrink the scans before a size-based split has to fight them.
Paper sizes
What the page dimensions in the parts list actually mean.