PDF Page Extractor, Explained
After reading this, you will know exactly what happens when you pull pages out of a PDF, how page numbering works, and how to build a clean new file with just the pages you need, in the order you want them.
What page extraction actually does
A PDF is a container. Inside it are page objects, and each page points at the fonts, text streams, vector paths and images it needs to draw itself. When you extract pages, the tool copies those page objects and everything they reference into a brand new PDF file. Nothing is re-rendered and nothing is flattened into a picture.
Here is the hook. Suppose you have a 40 page scanned contract and you only need pages 3, 4 and the signature page 39. Extraction produces a 3 page file where the text is still selectable, the vector logos stay crisp at any zoom, and the file is a fraction of the original size because the other 37 pages and the objects only they used are gone.
Contrast that with printing to PDF or exporting to images and back. Those paths rasterize the page, so a 300 DPI scan of a letter turns crisp black text into a grid of gray pixels. Extraction avoids that entirely.
Extraction is a copy operation, not a transformation. The bytes that describe each kept page are moved into the new file verbatim, which is why text stays searchable and quality never drops.
When to use it, and when not
Reach for extraction when you want a subset of an existing PDF as its own document: a single chapter from a manual, the two exhibits from a 90 page filing, or the pages of a bank statement you are willing to share while keeping the rest private.
Skip it when your goal is different. If you want every page as a separate file, that is a job for the PDF Splitter. If you want to stitch several PDFs together, use the PDF Merger. If the pages are sideways, fix orientation first with the PDF Rotator, then extract. And if your only problem is file size, extraction helps a little but the PDF Compressor is the direct fix.
How page numbering works
The single most common source of confusion is the gap between the number printed on the page and the number the tool uses. Extraction counts physical pages starting at 1, in the order they appear in the file. A book whose first printed page is labeled "1" often has a cover, a copyright page and a table of contents in front of it, so the printed page 1 might be the fifth physical page.
Here n_{\text{physical}} is the position the extractor uses, n_{\text{printed}} is the number on the paper, and the front matter count is how many unnumbered pages sit ahead of printed page 1. If a document has 4 front matter pages, printed page 10 is physical page 14. Always trust the thumbnails, not the ink.
Ranges and order, precisely
When you type ranges, three rules decide the output. First, a range like 2-5 means the four pages 2, 3, 4 and 5, inclusive on both ends. The count of pages in a range from a to b is:
So 2-5 gives 5 - 2 + 1 = 4 pages, not 3. The plus one trips up almost everyone once.
Second, order is preserved as written. Typing 5,1,3 produces a three page file in that exact sequence: old page 5 first, then old page 1, then old page 3. Extraction is also a reordering tool.
Third, repetition is allowed. Typing 1,1,2 gives you page 1 twice. That is occasionally useful for building a cover sheet, and occasionally an accident, so read your selection back before you build.
A descending range such as 5-2 is ambiguous. Some tools read it as empty, others as pages 5, 4, 3, 2. If you want reverse order, type the pages out: 5,4,3,2.
A worked example
Work through a concrete case so the numbers are checkable.
Extracting three pages from a 12 page report
Start with a report that has 12 physical pages: a title page, a one page summary, then 10 pages of body. You want the summary and the last two body pages, in that order.
- Map printed to physical. The summary is printed "page 1" but sits second physically, so it is physical page 2. The body prints "1" through "10" starting at physical page 3, so printed body page 9 is physical page 11 and printed body page 10 is physical page 12.
- Write the selection as physical pages:
2,11,12. That is three items, so the output has 3 pages. - Check the range math on the body pair. As a range it would be
11-12, which is 12 - 11 + 1 = 2 pages. Combined with page 2 that is 1 + 2 = 3 pages total, matching step 2. - Build the file. Page order in the result is 2, 11, 12: summary first, then the final two body pages.
The original file was 12 pages at, say, 480 KB. The output holds 3 pages. It is not exactly a quarter of the size, because shared resources like the embedded font get copied once, but it lands near 480 \times \frac{3}{12} \approx 120 KB plus a small fixed overhead.
Why file size does not scale exactly
You might expect keeping 3 of 12 pages to give a file exactly 25 percent as large. It rarely does, and the reason is shared objects. A font embedded once serves every page that uses it. A repeated header logo is stored once and referenced many times. The new file must keep a full copy of every object your chosen pages touch, plus a small structural overhead for the PDF header, cross reference table and trailer.
In words: the output is a fixed overhead of a few hundred bytes, plus the bytes unique to each kept page, plus one copy of the resources those pages share. If your three pages are heavy scans and the other nine are light text, the output can be larger than a quarter. If it is the reverse, it can be much smaller.
Common mistakes
Four errors account for most bad extractions.
- Off by one on ranges
- Typing
2-5and expecting 3 pages. It is 4 pages, because both endpoints are included: 5 - 2 + 1 = 4. - Trusting printed numbers
- Selecting printed "page 10" when front matter shifts it. If 4 unnumbered pages lead the file, printed 10 is physical 14. Use the thumbnails.
- Forgetting order equals output order
- Typing
3,1,2when you meant1,2,3. The result will genuinely start with the old page 3. - Assuming metadata leaves too
- Extraction copies the document, so title, author and other metadata usually travel into the new file. If you are sharing a subset for privacy, check that separately.
A new PDF built from a private original can still carry the original's author, creation software and title fields. Run the output through the PDF Metadata Viewer & Stripper before sending it to anyone.
Related tools
Extraction sits in a small family. Once you have your subset, you might feed it to the PDF to Images tool to get PNGs of just those pages, or go the other way with Images to PDF to assemble scans into a document you then trim. Each tool does one job, and chaining them keeps every step lossless except the ones that rasterize by design.
Frequently asked questions
Does extracting pages reduce quality?
No. The page objects are copied byte for byte. Text stays selectable and vector graphics stay sharp at any zoom. The only thing that shrinks is the file, because dropped pages and their exclusive resources are gone.
Can I extract pages in a different order than the original?
Yes. Order follows your selection exactly. Typing 4,1,2 puts the old page 4 first in the output. This makes the extractor double as a reorder tool for the pages you keep.
Why is my extracted file not exactly the fraction of pages I kept?
Shared resources are copied once regardless of how many pages use them, and every PDF carries a small fixed structural overhead. Keeping 3 of 12 pages that are text heavy can land well under a quarter; keeping 3 scan heavy pages can land above it.
Are my files uploaded anywhere?
No. The tool runs entirely in your browser. The PDF is read, the pages are copied, and the new file is created on your device. Nothing is sent to a server.
What is the difference between extracting and splitting?
Extraction produces one new file containing the pages you chose. Splitting produces many files, typically one page each or one file per range. Use the splitter when you want the whole document broken apart, not a single subset.