How do I write the page selection?

Numbers and ranges separated by commas, such as 1-3, 5, 8-10, with spaces around a hyphen allowed. Anything that is not a number or a range is refused and the offending part is quoted back, so a typo stops the run rather than quietly extracting the wrong pages.

Does the order I type change the order of the output?

Yes. Pages appear in the order you list them, so 5, 1 gives a two-page file that starts with page five. Listing a page twice does not produce it twice — repeats are removed — so the output page count is the number of distinct pages you named.

How large a PDF can I work with?

One file of up to 150 MB. Larger files are refused when you choose them. The page count is read first and shown on the page, so you can write the selection against the real number of pages rather than guessing.

Is the extracted file verified?

Yes. After saving it is opened again and its page count compared with how many pages you asked for. If the two do not match the run fails rather than offering a download, and the result panel shows the page count along with the input and output sizes.

How is this different from deleting pages?

They are the same operation described from opposite ends. Here you list what to keep; on the organise-pages tool you list the pages you want in the order you want them, and anything left out is dropped. Use this one when the pages you want are the short list, and that one when the pages you want to lose are the short list.

PDFPage management

Extract PDF pages

Choose pages or ranges, keep their requested order, and save them as a new PDF in your browser.

On-device prototype

Source PDF

File signature and page count checked locally

Scope: visible page content and order. Bookmarks, signatures, form behavior, attachments, document-level metadata, and certification state are not guaranteed in the new PDF.

About this PDF page extractor

To pull selected pages out of a PDF into a file of their own: choose the PDF, type the pages as numbers and ranges — 1-3, 5, 8-10 — and extract. The pages are copied whole into a new document in the order you listed them, and the file on your disk is not modified.

This is the keep-only-these half of page editing: whatever you list is what the new PDF contains, in your order. The document’s page count is read as soon as you choose it, so a number your file does not have is rejected before any work starts, with the valid range quoted. Duplicates are removed, and one range can cover at most 2,000 pages. Pages come across as pages — content, size and rotation unchanged — because they are copied rather than redrawn; it takes one PDF of up to 150 MB, and an encrypted file is refused with advice to remove its password locally first.

Runs in this tab — no uploadFree, no account, no watermark

Step by step

1Choose the PDFOne file of up to 150 MB. Its page count is read straight away and shown on the page, so you can write your selection against the real number of pages instead of guessing.
2Type the pages you want to keepNumbers and ranges separated by commas — 1-3, 5, 8-10. Spaces around a hyphen are fine. Anything that is not a number or a range is refused with the offending part quoted back, so a typo stops the run rather than quietly extracting the wrong pages.
3ExtractThe listed pages are copied whole into a new document in the order you wrote them. Nothing is redrawn, so page content, page size and rotation are exactly as they were.
4Download the new fileThe saved bytes are reopened and the page count compared with how many pages you asked for. A mismatch fails the run rather than offering a download; the panel shows the page count and the input and output sizes.

Splitting a PDF without a split button

There is no separate split control here, and that is a deliberate simplification rather than a missing feature. Every split anyone actually performs is a description of which pages to keep: the first chapter is 1-12, the appendix is 40-56, the one page you need to send someone is 7. Run the tool once per part and you have your split, with the exact page ranges written down in front of you rather than inferred from a number of parts.

The order you type is the order you get, so 5, 1 gives a two-page file that begins with page five. That makes this the quickest way to pull a small set of pages out in a specific sequence — an exhibit, a quote, a signature page and its annex — without opening a page editor.

What is preserved, and what is not

The pages themselves are copied, so their content, size and rotation are as they were and nothing is rasterised. Text stays selectable, vectors stay sharp, and an A3 page in the middle of an A4 document is still A3 in the output.

What does not follow is the material around the pages. The output is a newly created document rather than a trimmed copy of the original, so document-level metadata does not come across. Listing a page twice does not produce it twice either — repeats are removed — so the output page count is the number of distinct pages you named.

Limits and refusals

One PDF of up to 150 MB; larger files are refused when you choose them. A single range may cover at most 2,000 pages. A page number your document does not have is refused with the valid range quoted back at you, which is the useful form of that error — you learn the document is 46 pages long at the same moment you learn page 50 does not exist.

An encrypted PDF is refused with advice to remove its password locally and try again, rather than being partly read. Nothing read from inside a file ever appears in an error message.

Why the file stays with you

Every response from this site is served with a Content Security Policy whose connect-src directive is set to none. That is not a promise in marketing copy; it is an instruction to your browser, and the browser is the one enforcing it. While this page is open it cannot open a connection to anywhere — not to this site, not to anyone else — so there is no path by which your document could be uploaded, not by a bug, not by a future change, not by a script that should not be here. e2e/egress-proof.spec.ts asserts the served header and then tries to send data out by every route a page has and requires each attempt to be refused, and it runs on every build.

Extraction is often the step before sending something on — one page of a contract, the relevant month of a statement, a single certificate out of a bundle. It would be a strange privacy trade to upload the whole document in order to take one page out of it.

The second half of the guarantee is that the code has nowhere to send anything from. lib/tools/local-source-policy.test.ts reads every source file under lib/tools, workers, components and app on every test run and fails the build if any of them contains fetch, XMLHttpRequest, WebSocket, EventSource, sendBeacon or a peer connection, or even a remote address written down in a comment. A tool here cannot acquire an upload by accident, because the file that would have to contain one cannot be committed.

Questions people ask

How do I write the page selection?: Numbers and ranges separated by commas, such as 1-3, 5, 8-10, with spaces around a hyphen allowed. Anything that is not a number or a range is refused and the offending part is quoted back, so a typo stops the run rather than quietly extracting the wrong pages.
Does the order I type change the order of the output?: Yes. Pages appear in the order you list them, so 5, 1 gives a two-page file that starts with page five. Listing a page twice does not produce it twice — repeats are removed — so the output page count is the number of distinct pages you named.
Are the extracted pages altered in any way?: The pages themselves are copied, so page content, page size and rotation are as they were, and the file is saved with object streams. What does not follow is the material around the pages: the output is a newly created document rather than a trimmed copy of the original, so document-level metadata does not come across.
How large a PDF can I work with?: One file of up to 150 MB. Larger files are refused when you choose them. The page count is read first and shown on the page, so you can write the selection against the real number of pages rather than guessing.
Is the extracted file verified?: Yes. After saving it is opened again and its page count compared with how many pages you asked for. If the two do not match the run fails rather than offering a download, and the result panel shows the page count along with the input and output sizes.
How is this different from deleting pages?: They are the same operation described from opposite ends. Here you list what to keep; on the organise-pages tool you list the pages you want in the order you want them, and anything left out is dropped. Use this one when the pages you want are the short list, and that one when the pages you want to lose are the short list.