Split a PDF without uploading it: page ranges and shared resources

Splitting is usually one of two jobs: “I only need pages 3 to 7 for this email”, or “each page has to be a separate file for this portal”. Neither needs a server. Split PDF does both in the browser tab, and the original file on your disk is not changed.

How to split

1. Drop a PDF on Split PDF. Page thumbnails render in the tab.
2. Select pages by clicking thumbnails. All and None are shortcuts; start from None if you only want a few pages.
3. Choose the output. “One PDF per selected page” gives you separate files, packed in a ZIP when there is more than one. “One PDF with the selected pages” gives you a single shorter document.
4. Download.

To reorder pages, or to combine pieces of different files, use Merge PDF. It also lets you expand a file and exclude individual pages.

What splitting keeps

Split copies each selected page into a new file: its content, fonts, images, and page size. It does not resample images or convert text into pictures. A 300 ppi scan stays 300 ppi, selectable text stays selectable, and a Letter page stays a Letter page. Quality does not change.

What does not come along is document-level information. In our test output, the document properties held only a Producer entry (“UnlimitedPDF”), a Creator entry naming the pdf-lib library that builds the file, and creation dates. The source file’s title, author, subject, and keywords were not copied. Earlier saved versions of the document, which some editors leave in the file after an update, are not copied either, because only the current pages go into the new file.

Why a one-page extract can stay large

People expect page 1 of a 40-page file to be about one fortieth of the size. That only holds if every page carries its own images. PDFs often store an image once and reference it from many pages: a letterhead, a logo, a background texture, a banner.

We built a 40-page report where every page shows the same 4032×1176 banner photo, a 1.44 MB JPEG stored once in the file. The whole report was 1.49 MB. Then:

The page paints the banner, so the banner has to come along. Splitting the report into single pages multiplied the total size by about 40, because each file needs its own copy of the shared image. That is not a bug and it is not quality loss. It just means splitting is not a way to make things small. Compression is.

Split first, then compress

If the goal is a smaller attachment, the order matters:

1. Split out the pages you are sending.
2. Compress the extract.

Compressing first also works, but it spends time and memory re-encoding pages you are about to discard. On a phone, a large scan can run out of memory in the compressor when the three pages you need would have been easy.

Picking pages in long documents

• For a range like 12 to 30, select None, then click each page from 12 to 30. Scrolling through thumbnails is the slow part, not the split itself.
• For every page as its own file, keep All and choose “one PDF per selected page”. Expect a ZIP, and expect the total to grow if the pages share images, as in the table above.
• Check page numbers against the thumbnails, not against the numbers printed on the pages. A document whose first printed page number is “1” after a cover and a contents page starts at thumbnail 3.

Links and bookmarks

A link inside the document that points to a page you did not extract has nowhere to go. Bookmarks (the outline panel) are document-level structure and should be rebuilt if the extract needs them. For how this works in a merge, which uses the same page-copying method, see why merged PDFs look wrong; we tested bookmarks and internal links there.

File names

With “one PDF per selected page”, the files inside the ZIP are named after the source file with the page number appended. Our 40-page test file produced names ending in -p01.pdf through -p40.pdf. The zero padding keeps the files in page order when a file manager sorts them by name, so page 10 does not land between pages 1 and 2. Rename the ones you send if the recipient expects something more descriptive, such as a portal that wants “invoice-march.pdf” rather than a page number.

Why doing it locally matters

The file is read with the browser’s File API and processed in this tab. Thumbnails are rendered locally. Nothing is sent to a server, so splitting a medical record or a signed contract leaves no copy anywhere except your downloads folder. Closing the tab clears it from memory.

How we measured

Measured on 27 September 2026 with this site’s Split and Compress tools in Chrome on a Linux desktop. The 40-page test report was generated with ReportLab, and outputs were checked with pikepdf. Sizes are as the tool displays them (1 MB = 1,048,576 bytes), except the byte total, which is exact.