How to split a PDF and keep the text selectable

Split a PDF the wrong way and you get back a file that looks identical and behaves nothing like the original. You cannot select a word in it. Searching finds nothing. It is several times the size. Nothing downstream — no converter, no signature, no page stamp — will work on it properly again.

All of that follows from a single decision made inside whatever tool you used: were the pages you kept copied, or repainted?

Copied, or redrawn

A PDF page is not an image. It is a content stream — instructions describing text with its fonts and positions, vector paths, and sometimes an embedded photograph or two. That is why a PDF of a report is a few hundred kilobytes and still sharp at any zoom.

Copying a page moves that whole object into a new document. Every instruction that was in it is still in it, so the result is the same document with fewer pages.

Rendering a page runs those instructions onto a canvas and keeps the pixels. What you get back has the same appearance and none of the substance: the words are now a picture of words, the vectors are now anti-aliased edges, and the file is much larger because a page of prose compresses far better as text than as an image of text.

PDF Splitter copies. The panel says so plainly — pages are copied, not redrawn — and it is the most important sentence in the tool.

The pattern to watch for elsewhere is any workflow that renders and re-exports. Printing to PDF does it. So does any tool that shows you a rendered preview and then hands you back what it showed you.

Two libraries, doing two different halves

The tool uses two libraries and it is worth understanding why, because the split between them is the design.

pdf.js renders the thumbnails. Each page is painted onto a canvas at 132 pixels wide and turned into a JPEG. Those images exist for one purpose — so that you can recognise a page by looking at it — and they are thrown away afterwards.

pdf-lib copies the pages. When you press save, the pages you kept are lifted out of the original file as objects and placed into a new one. Nothing that was rendered ever reaches the output.

The reason the thumbnails are there at all: choosing pages by number out of a scanned contract is guesswork. You end up opening the file in a viewer, writing numbers on a scrap of paper, and coming back. Choosing them by looking is not guesswork, and it is faster.

Two details that keep that usable on a long document. The list appears before the thumbnails do, so a 300-page file is selectable immediately and the pictures fill in behind you rather than blocking the first paint. And thumbnails stop after 240 pages, past which they cost more memory than they are worth — the pages beyond that are all still there and can be picked by number.

Rotating rotates the page, not the view

A PDF page carries its own rotation, as a multiple of 90 degrees, and every reader in the world honours it.

Turning a page in this tool sets that value on the copied page. It does not bake the turn into pixels, because there are no pixels to bake it into — the page is still the same page, with the same text, simply declared to be the other way up.

Which means a sideways scan can be fixed permanently, in a file that is still a document afterwards. That is not true of most ways of straightening a page.

Picking the pages

Click a page to drop it; click it again to bring it back. A dropped page fades rather than vanishing, because it is still there and you may well change your mind.

Above the grid: All, None, Odd, Even and Invert — odd and even being the two that turn a duplex scan into two documents. Or type a range in the usual form, 1-3, 7, 12-15, and apply it.

The arrows under each page move it earlier or later, which quietly makes this a page-reordering tool as well as a splitting one.

Three ways to save:

• One PDF with the pages I picked — the common case.
• A separate PDF for every page picked — one file per page, zipped.
• Split into equal chunks — every n pages becomes a file, also zipped.

What it will not do

Open a protected file. A password-protected PDF has to be unlocked first, and PDF Protect & Unlock is where that happens.

Anything over 100 MB. Past that the thumbnails exhaust the tab, and the tool says so rather than freezing.

Compress. Copying pages keeps them exactly the size they were, which is the point — but it does mean a big document stays big. PDF Compressor is the separate job.

Merge. Taking pages out is this tool; putting files together is PDF Merger.

The rest of the toolbox

Around it: PDF Merger joins files, PDF Compressor shrinks them, PDF Watermark & Page Numbers stamps them and PDF Protect & Unlock handles passwords.

Out of the format: PDF to Word rebuilds a document as an editable .docx, PDF Text Extractor takes just the words, and PDF to Image does the rendering deliberately, when a picture of a page is what you actually wanted. Going in, Image to PDF makes one out of a pile of pictures.

They are all in the full tool directory, and every one of them runs on your own machine.