Two versions of a document in, a line-by-line diff out: what was added, what was removed, and how similar they are.
Contracts come back 'with small changes', policies get quietly revised, and manuscripts accumulate versions, this tool answers the only question that matters: what actually changed? Both documents are reduced to text (PDFs through the reader engine's text extraction, everything else read as text directly), split into lines, and compared with a longest-common-subsequence diff, the same algorithm family behind code review tools. Additions and removals are shown in order with the unchanged context around them, plus a similarity score.
Practical scope: the comparison is textual. Formatting changes, moved paragraphs (which show as remove+add pairs), and image changes are invisible to it; very long documents (beyond a few thousand lines) fall back to a summary mode listing lines unique to each side, because exact diffing grows quadratically. Word documents should be pasted as text or saved as .txt/PDF first, the .docx binary format is not read directly. Everything happens locally, so confidential drafts never leave the machine.
Why does a moved paragraph show as deleted and re-added? Line-based diffing has no concept of 'moved', it sees lines leave one place and arrive at another. The content is preserved; read remove/add pairs at different positions as a move.
What does the similarity score mean? The share of lines that are unchanged between the two versions (weighted as 2×unchanged ÷ total lines). 100% is identical text; below 90% on a 'lightly edited' document deserves a careful read.
Can it compare two PDFs with different layouts? Text order in PDFs follows the file's internal reading order, which usually matches visual order for born-digital PDFs. Scanned PDFs have no text layer, OCR them first with the OCR tool on this site.