You have two versions of something and want to know what changed. A contract that came back from the other side with “minor edits”. An article after an editor has been at it. Two exports of the same list from different days. A config file that worked yesterday and does not today. Reading both side by side and hunting for differences is slow and unreliable, because the eye skips right over a changed “not” or a moved comma. A diff does the hunting for you.
How to enter the two texts
There is one box, so the two versions are separated by a line that contains only three dashes:
The original text goes here.
---
The changed text goes here.The tool splits at that line and compares what comes before it to what comes after it. Lines made of three or more dashes or equals signs both work. If it cannot find a separator it will tell you.
Line mode and word mode
Line by line is the classic diff, the same idea that version control uses. Each line is either unchanged, removed, or added. Removed lines are shown with a minus sign and added lines with a plus. A line that was edited appears as a removal followed by an addition, because at this level a line is either identical or it is not. Untick “Show unchanged lines” to see only the changes, which is much shorter for large texts.
Word by word is what you want for prose, where the edits are inside sentences. Removed words are wrapped like this: [-old-], and new words like this: {+new+}. A changed sentence reads in one line, like The quick [-brown-]{+red+} fox. It is slower to read for large documents but much clearer for small edits.
The ignore settings
“Ignore extra spaces” is on by default, so a line that differs only in indentation or double spaces does not show as a change. That is usually right for prose and often wrong for code, where whitespace can matter. Untick it for those. “Ignore upper and lower case” is off by default, since a capital letter changing is often a real edit.
What it does well and what it does not
The comparison finds the longest run of lines or words the two versions have in common and reports the rest as changes, which is the standard approach. It handles added and deleted sections well. A paragraph that was moved shows up as a deletion in one place and an addition in another, since a basic diff has no notion of “moved”.
Very large texts are refused rather than allowed to freeze your browser; compare sections of the document instead.
Cleaning first
Differences that are only whitespace make a diff noisy. Remove extra spaces and remove empty lines on both versions first if that is a problem. If you are comparing lists rather than prose, compare two lists tells you what is in one and not the other without caring about order, and text similarity gives a single percentage for how alike two texts are.