Compare two PDFs

How to compare two PDF files

1. Open both files. Tap Original PDF for the older version and Changed PDF for the newer one, or drop a file on each box. ⇄ swaps them.
2. Read the changes. The text of both is shown side by side: words only in A are struck through, words only in B are marked. The count and the list say how many changes there are and on which page.
3. Step through them. Next and Prev (or a tap on a line in the list) scroll both columns to the same change.
4. Quiet the noise. Tick Ignore case or Ignore punctuation when only wording matters. Spacing and line breaks are ignored unless you untick that box.
5. Check the look. Visual compares each page pair as pictures and boxes what changed: signatures, stamps, images, bold or moved text. Use the slider or onion skin to see before and after.
6. Keep a record. Download changes report saves the list as an HTML file to attach to an email or print.

What it cannot do

• It compares words and pixels, not meaning. "30 days" and "thirty days" are a change; a rewritten clause that says the same thing is a change too. Reading what a change means is still your job.
• Formatting changes show only in Visual mode. The text compare sees words, so text made bold, a different font or a new colour is not a text change. Visual mode shows it as a changed area.
• Password-protected PDFs must be unlocked first. The page says when a file is locked; Unlock PDF removes the password if you know it.
• Scanned pages have no words to compare. A page that is only a picture is flagged; Visual mode still compares it, or get its words with Image to Text first.
• Visual mode compares page 1 with page 1. If an edit pushes text onto the next page, every page after it looks changed in Visual mode; the text compare does not have this problem.
• Text order comes from the PDF. Multi-column layouts, tables and footnotes are read in the order the file stores them, which can differ between two exports of the same document and then shows as moved words.

How it works

Text. pdf.js reads the text of every page. Both documents are cut into words and compared as one long sequence with the Myers diff algorithm, the shortest list of insertions and deletions that turns A into B (the same method git diff uses, here on words instead of lines). Neighbouring deletions and insertions form one change: "thirty" out and "sixty" in is one replacement. Because page breaks are not words, a sentence that slid from page 1 to page 2 is not reported.

Visual. Each page pair is drawn at the chosen resolution (100 dpi makes an A4 page 827 × 1169 pixels). A pixel counts as changed when its red, green or blue moves by more than the tolerance; the default 32 of 255 ignores the faint edge differences two PDF writers leave on the same text. Changed pixels close together share one box.

Questions

How do I compare two PDFs for differences?

Open the older version as A and the newer as B. The page reads the text of every page and lines the two documents up word by word: removed words are struck through in A, added words are marked in B, and the list says where each change is, for example "Page 3: thirty → sixty". Next and Prev step through the changes. Visual mode then compares each page pair as pictures and boxes the areas that look different, which also catches changed images, signatures and formatting.

Can I compare PDF files online for free without uploading them?

Yes. This page is free with no account and no limit, and it never uploads the files: pdf.js, the PDF reader built into Firefox, runs inside this page and reads both files from your device. That matters for the documents people most often compare, such as contracts, leases and offers, which you may not be allowed to hand to a third-party server. Very long files are limited only by your device's memory.

Can it compare two Word documents or scanned documents?

Only as PDFs. Save or export each Word document as PDF (File → Save as → PDF) and compare the two PDFs here; the text compare works on the words, so differences in layout between the two exports do not show as changes. A scanned document has no text layer: the page says which pages have none, and Visual mode still compares them as pictures. To get words out of a scan, run it through Image to Text first. For two pieces of plain text, paste them into Diff Checker instead.

Why are some changes shown on two pages?

The text compare treats each document as one long run of words, so a sentence that moved from the end of page 1 to the top of page 2 is not a change. When an edit pushes text onto another page, the change is listed with the page in A and, if it differs, the page in B, for example "Page 4 (B: page 5)". Visual mode compares page 1 with page 1, so reflowed text shows there as changed everywhere below the edit.