Extract PDF pages

Choose a PDF, type the pages you want, and get one new PDF containing only those pages. The page objects are copied across untouched, so the text stays selectable, the images stay at their original quality, and nothing is re-encoded.

Extracting is not the same as splitting

Splitting a PDF writes every page out as its own file — twelve pages become twelve files, and you then delete the eleven you did not want. That is the right tool when you genuinely need each page separately.

Extracting keeps your selection together in one document, in reading order. That is what you want for the three pages of a contract that actually matter, the two pages of a bank statement showing one transaction, or chapter four of a manual. One file, right pages, nothing to tidy up afterwards.

How to write the page numbers

Single pages are separated by commas, and ranges use a hyphen: 1, 4, 9-12 gives you six pages. Order does not matter — the output is always in document order, so 9-12, 1 produces the same file as 1, 9-12. Duplicates are collapsed, so 3, 3, 3 selects page three once.

Open-ended ranges work too. 5- means "page five to the end", and -3 means "the first three pages". As you type, the line under the controls tells you exactly how many pages will be kept and which ones, so you can check before committing.

Full-width punctuation is accepted. If you are typing on a Chinese input method, the comma you produce is a full-width one and the dash may be a full-width hyphen or a wave dash. Most English-language tools reject the whole string when they see those, which looks like the user made a mistake when in fact the tool simply did not handle it. Here they are normalised before parsing.

What is preserved and what is not

Page content is copied as-is. Text remains real text you can select and search, vector graphics stay vector, and embedded images keep their original compression — extracting never re-encodes anything, so there is no generational quality loss no matter how many times you do it.

Some document-level structures do not survive, and it is better to know in advance. Bookmarks and the outline tree refer to pages by object reference, so entries pointing at pages you did not keep are dropped. Form fields are copied, but scripts and calculations that reference other pages may no longer work. Internal links to removed pages become dead. Attachments, document-level JavaScript and some metadata are not carried over.

File size will usually not shrink in proportion to the page count. Fonts and images live in a shared pool, and anything referenced by a page you kept has to be copied in full. Extracting two pages from a 60-page report that uses one large embedded font can produce a file far larger than one-thirtieth of the original.

Why doing this locally matters more than usual

Think about what people actually extract pages from: a signed lease, a medical report, a payslip, a passport scan, a tax filing. The reason for extracting is normally that the rest of the document is information you specifically do not want to send. Uploading the whole file to a server in order to remove the sensitive pages defeats the purpose in the most direct way possible.

Here the file is parsed by pdf-lib inside this browser tab and the new document is written to a Blob your browser downloads. There is no request carrying the original, so the pages you excluded were never transmitted either.