Merge PDF Files Online

• Split PDF
• Compress PDF
• Rotate PDF
• Reorder PDF pages
• Remove PDF pages
• Add page numbers
• Watermark PDF
• Edit PDF metadata
• Images to PDF
• PDF to JPG

A single document is easier to send, easier to file and easier to read than a folder of fragments. Contracts arrive as a cover letter, a body and a signature page. Expenses arrive as thirty phone photographs of receipts. A tender response is assembled from four departments and has to reach the buyer as one file with one page numbering. Merging is the step that turns a pile into a document, and the difference between a good merger and a poor one is entirely in what survives the process.

What a merge should preserve

• Selectable text. Pages are copied as objects, so the text layer, the embedded fonts and any tagging arrive intact. A merger that rasterises pages turns a searchable contract into a picture of a contract.
• Original image quality. Nothing is re-compressed. A 600 dpi scan is still a 600 dpi scan in the merged file, and a vector chart is still vector.
• Each page's own size. A4 body pages and a Letter appendix keep their own dimensions rather than being stretched onto a common sheet.
• Links and form fields. Internal structure travels with the pages instead of being flattened away.
• A way to navigate. A merged document needs an outline, or the reader is left scrolling through three hundred pages looking for section four.

Every file you add is read once, on your device, to find its page count and the size of each page. Those numbers drive everything the rail shows you — the running page total, the range field's bounds, the divider size preview — so the figure on screen before you press Merge is the figure the merge produces. When you press it, the sources are handed to a background worker, the pages you asked for are copied into a new document in the order you set, and the finished file comes back as a preview you can check before downloading.

What happens when you press Merge

• Sources are handed over, not copied. Each document's bytes are transferred to the background worker rather than duplicated, so a queue of large scans does not need twice its own size in memory.
• Pages are copied per document, not per page. Copying a document's pages in one operation means its shared resources — fonts, images, colour spaces — are stored once in the merged file rather than once per page.
• Rotation is written, not rendered. A rotation becomes an attribute on the page. Nothing is rasterised, so a rotated scan is byte-for-byte the same image it was.
• The outline is built last. Once every page is placed, the bookmark for each source document is pointed at the page index it actually landed on.
• The file is written with a compressed cross-reference. Object streams keep the structural overhead of a fifty-file merge small instead of linear in the number of documents.

Clicking page thumbnails is fine for a five-page file and unusable for a three-hundred-page one. Every file in the queue therefore has a page field that takes a range expression, and it is the fastest control on the page once you know the four forms it accepts.

The forms a page field accepts

• Empty. Every page of that file, which is what you want most of the time.
• 3 — a single page.
• 1-3,7,12- — a list. Spans are inclusive at both ends, an open end runs to the last page, and an open start begins at page one, so -5 is the first five pages.
• 1,1,2 — a repeat. Listing a page twice copies it twice, which is how you put the same cover sheet in front of several sections.
• 5-1 — a descending span, which includes those five pages in reverse order.

What happens when a range is wrong

A page number the file does not have is refused, not quietly clamped. Asking for 1-50 from a twelve-page document does not hand you twelve pages and hope you do not notice — the card says the file only has twelve pages and the merge waits until you fix it. The same applies to page 0, which does not exist in any PDF, and to anything that is not a number. A wrong answer that looks right is worse than an error message, and this is the control most likely to produce one.

The order of the queue is the order of the merged document, and there are three ways to set it: move an individual file up or down, sort the whole queue, or reverse whatever order you have. Sorting is the one worth explaining, because the obvious implementation of it is wrong.

Sorting a folder of scans

• By filename, naturally. A plain alphabetical sort puts page10 before page2, because it compares character by character and 1 sorts before 2. Numbered scans are the single most common thing anyone merges, so filename sorting here compares the numbers as numbers.
• By size. Useful when the odd one out is the thing you are looking for — the 40 MB scan among thirty 400 KB pages is usually the page that was photographed rather than scanned.
• By date modified. The order the pages were captured in, which for a batch of phone photographs is usually the order you want.
• Reverse. An independent switch rather than six combined options, so it applies to whichever sort you chose.

Rotating a document as it merges

Each file has a rotation of 0, 90, 180 or 270 degrees. The angle is added to whatever rotation the page already carried, so a page stored sideways at 270 degrees plus another 90 comes out upright at 0 rather than at 90 — which is what the button says it will do and what a scan of a landscape page actually needs. Because rotation is an attribute rather than a redraw, applying it costs nothing and changes no pixel.

The merged document is written with a navigation outline: one top-level bookmark per source file, named after the file, pointing at the first page that file contributed. Open the result in any reader and the navigation pane lists your source documents; click one and it jumps there.

Why this matters more than it sounds

• A merged file loses its filenames. Once thirty documents are one document, the only remaining trace of where each began is the outline. Without it, the structure of the thing you assembled is gone.
• Nobody else does it. The large free web mergers produce a flat document with no outline at all, which is why a merged tender response is so awkward to review.
• Duplicate names are disambiguated. Two files called scan.pdf become two distinct bookmarks rather than two identical entries you cannot tell apart.
• Non-Latin names survive. Titles are written as Unicode, so Japanese, Arabic and Cyrillic filenames appear correctly instead of as mangled Latin-1.
• It can be turned off. If you are merging fragments of one document rather than several documents, an outline listing them is noise. One switch removes it.

Nothing is resized. An A4 report, a Letter appendix and a landscape spreadsheet export merge into one file in which each page keeps its own dimensions, exactly as a professional layout would. Readers handle this without complaint; forcing everything onto one sheet size would crop or letterbox real content.

Divider pages

• Between documents only. Two files produce one divider, five produce four, and one produces none. A divider after the last document would be a blank final page nobody wants.
• Match. The divider copies the size and orientation of the page before it, including a rotation — so a divider following a sideways landscape scan is landscape, not portrait.
• A4 or Letter. A fixed size instead, for a document that will be printed on one stock and where a mixed-size divider would jam a duplex feed.
• They count. Divider pages are included in the page total shown beside the Merge button and in the free page allowance, so the number you see is the number you get.

Being precise about this is more useful than a blanket promise of quality. A merge copies page objects between documents; that operation preserves a great deal and there are a few things it cannot carry.

Carried over

• Text with its embedded fonts, still selectable and searchable
• Images at their original resolution and compression — nothing is re-encoded
• Vector graphics as vectors
• Each page's own media box, orientation and rotation
• Annotations and links that belong to a copied page
• Colour spaces and transparency groups referenced by the page

Not carried over

• Encryption — an encrypted source cannot be opened at all, so it is refused up front
• Document-level structures that belong to a whole file rather than a page, such as a source document's own outline, which is replaced by the new one
• Attachments and JavaScript actions held at the document level rather than on a page
• The source documents' metadata, since a merged file needs its own

Free is not a demonstration. Page ranges, rotation, sorting, divider pages, bookmarks and reproducible output are all free, because they are what the tool is. What Pro raises is scale, and the reason is honest: every one of these limits exists to keep a first merge on a mid-range phone quick rather than to withhold a feature.

• Free. Up to 10 PDFs, each up to 25 MB, merged into a document of up to 100 pages.
• Pro. Up to 50 PDFs, no page limit, and no imposed size cap — the ceiling becomes what your own device can handle, which the tool estimates and tells you rather than guessing on your behalf.
• Pro also adds document details. Title, author, subject and keywords written into the merged file, so it arrives correctly labelled instead of needing a second tool afterwards.
• Over a limit, you are told which one. A refused file names its own size, the cap it exceeded and a link to the PDF compressor. Nothing reaches a progress bar and then fails.

• Contracts and signature pages. A body document plus a returned signature page, in that order, with a bookmark for each so the counterparty can find the signature.
• Expense claims. Thirty receipt scans sorted by date modified, which is the order they were photographed in, merged into one file for a claim form.
• Tender and grant responses. Sections from four teams, ordered by filename, each with its own bookmark and a divider page between them.
• Course and reading packs. Selected chapters from several sources — chapter three of one, the appendix of another — assembled with page ranges rather than by splitting each file first.
• Scanned books and reports. A batch scanner that produces one file per ten pages, sorted naturally so part 2 precedes part 10.
• Portfolios. A cover page repeated in front of each section using a repeated page range, so one file does the work of several.

• A file says it is protected. It is encrypted, and an encrypted document cannot be read without its password. Open it in your PDF reader, save an unprotected copy, and merge that.
• A file says it cannot be read. The file is damaged or is not really a PDF — a renamed image or a truncated download. Re-download it, or open and re-save it in a reader, which usually rebuilds the structure.
• The Merge button will not enable. It needs at least two readable files with no unresolved page-range errors, and a page total inside your plan's allowance. The line above the button says which of those is not yet true.
• The page total is higher than expected. Divider pages count, and a repeated page range counts once per repeat. Both are shown in the total before you merge.
• A rotated file still looks wrong. Rotation adds to the angle the page already has. If a page was stored at 180 degrees and still looks upside down after another 180, it was stored upright and the content itself is inverted — that needs the rotate tool, not the merger.
• The result opens without a navigation pane. Some readers keep the pane collapsed by default. The bookmarks are in the file; the pane is a view setting in the reader.

There is no upload step. The documents are read, merged and written where they already are, and the finished file is handed straight back to you. Nothing is transmitted, nothing is stored, and no copy exists anywhere for anyone to delete later — which is a stronger promise than a retention policy, because there is nothing being retained.

What follows from that

• It starts immediately. There is no upload to wait for, so merging begins the moment you press the button rather than after the slowest file has finished travelling.
• It works offline. Once the page has loaded, a merge needs no network at all. Disconnect and it still works.
• Confidential documents stay confidential. Contracts, medical records and financial statements are exactly the documents people most often need to merge, and exactly the ones they should not be posting to a stranger's server.
• Your device is the only limit. Because there is no server paying for bandwidth and storage per file, there is no commercial reason for a size cap beyond what your own machine can manage.

Pages are copied as document objects rather than redrawn, and the work runs in the background so the page keeps responding while a long merge finishes.