Merge scanned documents into one PDF

Merge first, compress second

Open the merge tool, add every scanned document, put them in order and merge. Then take the result to the compression tool and name the size you need.

The reverse order costs quality for no benefit. Compressing each part and then merging puts every page through image encoding, then copies the already encoded images into a new document, and you end up guessing at per-file sizes that add up to something unpredictable. One compression pass on the finished document targets the number you actually need.

Why the merged file is so big

A scan is a picture of a page. Even at a modest resolution, a picture of a page carries far more data than the same page as text, and a document feeder set to colour makes it several times worse again. Ten scanned documents of twenty pages each is easily a few hundred megabytes.

Merging does not add to that and does not take anything away. Pages are copied as they are, so the merged file is close to the sum of its parts.

How far a scan will compress

Scans are the best case for compression. A photographed page carries slack that a page of real text does not - pixels can be re-encoded smaller while the words stay the same - so a scan usually has further to fall than a born-digital document of the same size.

The engine works to a byte target rather than a quality slider. You name a size, several methods run against it, and each result is read back and checked before it is offered. No success figure is promised here: the tool reports the size it reached even when it misses, and a photographed page that is mostly ink may still land above your number.

What the merge keeps and what it clears

Every page keeps its content, its width and its height, so a stack that mixes A4 and Letter stays mixed. The merged file is reopened and compared page by page against receipts taken from the sources before it is handed over, and a mismatch means the merge is refused rather than downloaded.

Document properties are cleared, so the merged PDF carries no title, author or dates from any of the scans. Bookmarks, form fields, page labels and attachments do not survive either, and the tool names which sources carry them before it runs.

A scan that has been through text recognition keeps that text layer, because the text is part of the page. Its accuracy is whatever the recognition produced, which merging neither improves nor damages.

What the ranking pages cover

The ranking pages for this query teach capture first and merging second. WikiHow's page walks scanner, iPhone and Android capture before any merging step; the niche tools - AskPDF, PDFree, FileSplice - sell the merge itself as an upload. Those are their pages' claims, not measurements made here.

This guide starts where those pages stop: you already hold the scans and the question is order. Merge first and compress second, because one compression pass on the finished document targets the number you actually need, while compressing parts first guesses at per-file sizes that add up unpredictably. The merge clears properties, bookmarks, fields and labels and says so before it runs; the compression then reports the size it reached even when it misses.

Before you merge

Check that every part is the document you think it is. Scanner output named scan0042 tells you nothing, and finding a wrong file after merging means doing the whole job again.

If the parts arrived as photographs rather than PDFs, the multi-page scanning guide covers turning them into pages first. The merge hub lists the rest of the cluster.

Getting the order right across separate scans

Scanned files carry no clue about where they belong. A folder of scan0031 through scan0074 is in the order the sheets went through the feeder, which is not page order if the stack jammed, if somebody scanned the reverse sides in a second pass, or if one document was rescanned.

Open the first page of each file before you merge and write the real sequence down. Then set that order in the tool by dragging, and check the page count shown beside each file against what you expect. A file holding twelve pages where you expected one is worth catching before you merge rather than after.