This free extract pdf pages works instantly in your browser — no sign-up, no downloads. Answer: Open the PDF, type the pages you want, like 2,5-7 or 4-end, and download a new file with just those pages in your chosen order. Pages are copied, not redrawn, so text stays searchable. Bookmarks don't come along.
Type 2,5-7 — get a new PDF with exactly those pages
| You type | Pages in the output | Count |
|---|---|---|
4 | 4 | 1 page |
2,5,9 | 2, 5, 9 | 3 pages |
5-7 | 5, 6, 7 | 3 pages |
2,5-7,20 | 2, 5, 6, 7, 20 | 5 pages |
4-end (24-page file) | 4 through 24 | 21 pages |
1,1-3 | 1, 2, 3 — duplicates collapse | 3 pages |
20,5-7,2 (as typed) | 20, 5, 6, 7, 2 — your order, kept | 5 pages |
Ranges include both endpoints, "end" (or "last") stands for the final page number, and whitespace is ignored — 2, 5-7 and 2,5-7 are the same input. Every resolved list is shown as chips before you download, so there are no surprises.
Extracting is the polite sibling of splitting: you point at a set of pages, and the tool builds one new document that contains copies of those pages and nothing else. Under the hood it parses your selection into an exact page list, copies the chosen page objects into a fresh PDF with pdf-lib, and hands you the result. Because pages are copied rather than re-rendered, text stays selectable, images keep their resolution, and links and comments on the chosen pages travel with them. Bookmarks and form logic don't, as the tests below show.
Your input is split on commas; each entry is a number, a range like 5-7, or a range ending in end. Ranges are inclusive, duplicates collapse, and the validator rejects the three classics before any work happens: pages before 1, pages past the end of the document, and backwards ranges like 7-5. If the box says 25 on a 24-page file, you'll see exactly that error — not a wrong download.
Open the PDF (the page count appears immediately), type the selection, and read the chips: they show the precise resolved list, so "5-7" visibly becomes 5, 6, 7. Leave the order box unticked to keep your typed order — handy for pulling a summary to the front — or tick it to keep the document's own sequence. The download appends -extracted to your filename.
A 24-page quarterly report needs to go to someone who should see only the summary and two exhibits. Type 2,5-7,20: the parser resolves it to pages 2, 5, 6, 7, 20 — five pages — and the download is a 5-page PDF containing exactly those, in that order. The original 24-page file sits untouched on disk, and searching the extract finds nothing from the 19 pages left out. Want the exhibits in front instead? Type 20,2,5-7 and the output runs 20, 2, 5, 6, 7.
Three related jobs live next door: splitting when you want many files, deleting pages when it's easier to name what goes, and merging when the extract needs company.
You'd expect one page out of twelve to be about a twelfth of the size. Often it isn't. Pages share resources, mostly embedded fonts and sometimes a logo or background image, and every page that uses a font needs the whole font file copied with it. We ran this page's copy method over real PDFs and measured the output against the original:
| Document | Original | Page 1 alone | First half of the pages |
|---|---|---|---|
| 12-page text report | 154 KB | 79% | 88% |
| 14-page PDF guide | 69 KB | 48% | 76% |
| 66-page playbook | 429 KB | 69% | 84% |
| 49-page spec | 228 KB | 9% | 30% |
| 48-page product manual | 4.3 MB | 3% | 54% |
| 148-page brand guide | 5.0 MB | 0.5% | 67% |
| 206-page web archive | 2.7 MB | 4% | 81% |
Text-heavy documents with a few embedded fonts keep most of their weight even as a single page: page 1 of the 12-page report still came out at 79% of the full file. Image-heavy files behave the way you'd hope, because each picture belongs to one page. Page 1 of the 148-page brand guide was 23 KB out of 5.0 MB.
So if you're extracting pages to get under an email attachment limit, check the size before you send. When it's still too big, run the extract through the PDF compressor, or for scans, through its rerender mode.
Extraction builds a brand-new document and copies page objects into it. Anything stored on a page travels with it. Anything stored at the document level stays behind. We checked each item on the output files:
| Part of the PDF | In the extract | Why |
|---|---|---|
| Text and images | Copied | Pages aren't re-rendered, so quality is unchanged |
| Links on the page | Copied (14 of 14 on our test page) | They belong to the page |
| Comments and highlights | Copied | Also stored on the page |
| Bookmarks | Dropped on all 7 test files that had them | They belong to the document, not the pages |
| Fillable form | Form dictionary dropped | Fill or flatten the form before extracting |
| Password-protected source | Output comes out broken | Remove the password first |
The form row matters if you're pulling one page out of a long application. The field boxes still show, but the extracted file doesn't carry the document-level form, so in our test it no longer registered as fillable. Fill the form first, or flatten it so the answers become part of the page, then extract.
The password row is a real trap. A file with only a permissions password opens normally here, and the page count and selection work as usual, but the copied pages come out encrypted-yet-unreadable and other apps report errors or blank pages. We reproduced it with a 3-page test file. If your source has any password, remove it with the PDF password tool before extracting.
They all produce smaller files, and the choice is about how you'd rather describe the job. If you can name the pages you want, extract. If it's easier to name the ones to throw away (say, every page but the blank separators), delete pages instead. If you want several output files, one per chapter or one per invoice, split the PDF. All three copy pages the same way, so the size and bookmark notes above apply to each.
Open the PDF in an extractor, type the pages you want — single numbers, comma-separated lists, and ranges like 5-7 all work, and "end" means the last page — then download the new file it builds. This tool runs in your browser, so there's no upload, no account, and no watermark.
It extracts pages 2, 5, 6, and 7 into a new PDF. Commas separate entries, and a hyphen range is inclusive on both ends. The status line always shows the exact resolved list — here 4 pages — before you download anything.
Extracting keeps one subset as a single new file: 24 pages in, pages 2,5-7,20 out, one PDF. Splitting cuts the document into multiple files by range, chunk size, or one file per page. If you want exactly one smaller document, extract; if you want many outputs, split.
Yes. By default the output follows the order you typed — 20,5-7,2 produces a PDF whose pages run 20, 5, 6, 7, 2. If you'd rather keep the document's own order, tick "keep document order" and any selection comes out sorted.
No. The pages you chose are copied into a brand-new document and downloaded alongside the original, which stays on your disk untouched. The extracted text stays selectable and searchable — pages are copied, not re-rendered.
Deliberate repeats are collapsed (1,1-3 resolves to 1, 2, 3) to protect you from accidental double-typed entries. If you genuinely need a page twice — say a cover at front and back — run the extractor once for the body and then prepend the cover with a merge tool.
Pages share embedded fonts and images, and each extracted page needs the full font files it uses. In our tests a single page of a 12-page text report kept 79% of the file size, while one page of an image-heavy 148-page guide was 0.5%.
The original file is probably password-protected, even if it opens without a password. Its pages are copied still encrypted, so the new file can't be read. Remove the password from the original first, then extract again.