Merge PDF
Drop two or more PDFs and get one file back, in this tab. Each page is copied out of its source and written into a new document, so the pages keep their own size and rotation — and the copy leaves behind what lives at document level rather than on a page: bookmarks, form fields and the file's own title and author. The panel reports the page count, the page sizes it kept and the output size against the inputs. Nothing is uploaded.
Drop PDFs here
or choose files · two or more, up to 30 · 100 MB in total · or paste with Ctrl + V
PDF in, one PDF out · up to 30 files and 100 MB · page order is yours · page sizes and rotation kept · no upload · no account
Merge PDF: What a Page Copy Keeps, and What It Leaves Behind
Five PDFs went in with 29 pages and 3,999,881 bytes between them. One file came out at 3,597,920 bytes, 10.0% smaller than its inputs, and page 1 of the 15-page paper rendered to 1,090,584 pixels with none of them changed. The same paper's 22 bookmark entries did not come along, and neither did the single form field or the title and author lines of the other files tested: those live at document level, and a page copy never visits them.
A merge walks the page tree, not the document
A PDF is two things stacked together: a tree of page objects, and a set of document-level tables that point at those pages. The outline, the form, the name tree, the page labels and the metadata are all in the second group. A merger opens each source, reads its page objects and adds them to a fresh document. The fresh document starts with an empty catalog, so the tables are never carried over.
The measurements follow that structure. The 15-page paper carries 22 bookmark entries across its outline; the merged file has 0. A 300 × 300 test page built with one text field had 1 field before the merge and 0 after, and the field's value went with it. A document whose title read "Quarterly report" and whose author read "A. Person" produced a merged file where both were gone, and the producer and creator lines had been replaced with the library's own name.
That is not a bug to work around. It is what "merge" means at the file level: the pages move, the document around them does not.
The pixels and the words come through untouched
Page 1 of the paper was rendered before and after the merge at the same scale, 1.5×, which gives a 918 × 1,188 canvas. The two renders held 1,090,584 px each and 0 of them differed; the largest channel difference was 0.
The text layer survived the same way. Extracting the characters from the source page gave 2,802 characters, and the merged page gave 2,802, identical strings and the same SHA-256 prefix. Nothing was rasterized, so the words are still selectable, searchable and copyable in the output.
The size is neither the sum nor the largest input
| What was merged | Input bytes | Output bytes | Change |
|---|---|---|---|
| the 15-page paper, alone | 2,215,244 | 1,817,653 | −17.9% |
| paper + 7-page A4 set | 2,623,115 | 2,224,105 | −15.2% |
| paper + the same paper again | 4,430,488 | 3,639,328 | −17.9% |
| 3-page note + the same note again | 10,634 | 9,653 | −9.2% |
| five files, 29 pages | 3,999,881 | 3,597,920 | −10.0% |
The output is written from the pages, not by gluing the input bytes end to end. Each page object and the objects it references are copied into a new document, and the source's own cross-reference streams and document-level tables are left out. The result can be smaller than the sum and even smaller than the largest input, as the first and last rows show, and the five-file merge completed in 565 ms.
Do not read the smaller number as compression. The page content is byte-for-byte the same material; what disappeared is file furniture that was never part of any page.
Page size travels with each page
A merged file is not forced onto one paper size. Merging a 3-page A4 set with a 7-page A4 set gave a 10-page file whose pages alternate between 595.28 × 841.89 pt and 841.89 × 595.28 pt, exactly the orientation each page had in its source. A page built at 400 × 300 pt and rotated 90 degrees came out with its rotation intact, so it still displays as 300 × 400.
The panel reports the sizes it kept, grouped, in points. A merge of Letter and A4 documents reads back as "612 × 792 pt × 15 · 595 × 842 pt × 7 · 842 × 595 pt × 7" rather than a single blended dimension.
What the copy cannot carry
- Bookmarks. 22 entries in, 0 out. The outline tree belongs to the source document, and no page points back at it.
- Form fields. 1 in, 0 out, values included. The field widgets sit on pages, but the form dictionary that gives them names and types is a document-level object.
- Metadata. Title, author, subject and keywords are not copied. Producer and creator are written fresh.
- Page labels. None are written, so a source numbered i, ii, iii loses that numbering.
- Digital signatures. A signature is computed over the bytes of the document it signed. The output is a different file with different bytes, so no signature from a source page still covers it.
- Encrypted files. A PDF whose trailer carries an /Encrypt entry is refused before a page is read. The panel names the file and merges the rest.
What this page will not do
- Split, delete or reorder within a file. The unit is the whole file; its pages stay together and in order.
- Compress. A merge changes the container, not the page images or fonts.
- Read text or OCR anything. No text is extracted here; the words are copied, not recognized.
- Remove a password. An encrypted source is refused, not unlocked.
- Repair a damaged file. A PDF that cannot be parsed is reported by name and skipped.
- Sign or flatten. No signature is applied and no form is flattened.
- Take an unlimited pile. The limits are 30 files, 100 MB in total and 2,000 pages, because every byte stays in this tab's memory.
How to use this page
- Drop two or more PDFs onto the panel, or paste them with Ctrl + V.
- Put them in order with the up and down buttons; the list reads top to bottom like the output.
- The merge runs on its own after each change, and the result block reports page count, size and the page sizes kept.
- Download the merged PDF, or add more files and reorder again before downloading.
What the controls do
Sources is the list of files, numbered in output order. Each row shows the file name, its page count, its size and the page sizes inside it.
The arrow buttons move a file one place up or down; the first row cannot move up and the last cannot move down. The × button removes a file from the merge without touching it on disk.
Add more PDFs opens the picker again; Clear all empties the list and throws the built file away.
The result block reports the merged page count, the output size, the difference against the sources and every page size in the output, grouped by count.
Why this page doesn't upload your file
The PDF library is served from this site and runs inside the tab. Reading a PDF, copying pages and writing the new file all happen in the browser's own memory, so no request carries a document out. The page can show it: with the network cut after the library loaded, the same five files merged again to 29 pages and 3,597,920 bytes, with the same page sizes. Seven bytes separated that file from the online run, all of them inside the compressed streams.
Questions that come up
Does merging PDFs make the pages blurrier? No. The pages are copied as objects, not redrawn: page 1 of a 15-page paper came out of a merge with 0 of its 1,090,584 pixels changed, and its 2,802-character text layer identical. A tool that renders pages to images is compressing, not merging.
Why is my merged PDF smaller than the files I added? The new file is written from the pages. The source's cross-reference streams and document-level tables are not copied, and the paper merged on its own came out 17.9% smaller. The page content itself is unchanged.
Why did the bookmarks and form fields disappear? They are document-level structures, and a page copy does not visit them. The 22 bookmarks in the test paper became 0 and a text field became 0. If a filled form has to stay fillable, keep the original and send the merged copy as a read-only version.
Can I merge a PDF that asks for a password? Not here. A file with an /Encrypt entry is refused before a page is read, because its pages cannot be decoded without the key. Open it once in the app or tool that owns the password, export an unprotected copy, and merge that.
Can I merge PDFs on a phone? Yes, within memory: the 29-page merge used the tab's own memory for 4.00 MB of input. Large scanned files are the practical limit, which is why the ceiling is 100 MB across 30 files.