Text Compare

How does the text comparison work?

The tool finds the longest run of words the two texts have in common, in order, and marks everything else as removed from the first text or added in the second. That shared run is called the longest common subsequence, and finding it is what every diff checker does; this one uses Myers' algorithm, the same method behind git diff, in a form that needs memory only in proportion to the length of the texts.

Before comparing, both texts are cut into pieces. In Words mode a piece is a word, a punctuation mark or a run of spaces. In Lines mode it is a whole line, and every changed pair of lines is then compared word by word so you can see what moved inside the line. In Characters mode the texts are compared word by word first, and then each replaced stretch of up to 4,000 characters is compared again letter by letter — that is what turns a typo into one highlighted letter instead of a whole highlighted word. Chinese, Japanese, Thai, Lao, Khmer and Burmese, which are written without spaces, are compared one character at a time.

The comparison runs in a background thread of your browser, so a long text does not freeze the page. Each text can be up to 300,000 characters. Two unrelated texts of that size could take minutes to align exactly, so the work stops after five seconds and the remaining differences are shown as whole blocks; the page says when that happens.

What do the ignore options change?

Ignored characters are still shown in the result. They just never count as added, removed or changed.

What the numbers mean

The counts are in the unit you compare by. Added is what appears only in the second text, removed is what appears only in the first, and changed is what was replaced in the same place, counted on the longer side. In Words mode a punctuation mark counts as one word. The same is the share of both texts that matched: twice the unchanged part divided by the length of both texts together, so two identical texts score 100% and two texts with nothing in common score 0%.

"Hide unchanged text" keeps 160 characters, or three lines, around every change and folds the rest, which makes a long document with a few edits easy to scan. "Copy result" gives you the comparison as plain text: [-removed-] and {+added+} marks inside the text, or lines starting with - and + in Lines mode.

Where your texts go

Nowhere. Both texts are compared on your device; nothing is sent to a server or saved, so contracts, drafts and student work are safe to paste. You can also open a file: plain text files up to 5 MB, and Word .docx files, whose text is read in the browser.

Limits worth knowing

A diff shows what changed, not why: a moved paragraph appears as removed in one place and added in another. Formatting in Word files — bold, tables, comments — is not compared, only the words. And on two texts that share almost nothing, the five-second budget means some differences are shown as larger blocks than strictly necessary.