Two versions of the same document, and somewhere inside them a sentence changed: a price, a date, a “shall” that became a “may”. Reading both side by side to find it is slow, unreliable work. This tool extracts the text of both PDFs and diffs them word by word — removed words struck through in pink, added words in green — with page counts, word counts and a similarity figure up front, all computed on your device.
How the comparison works
Each document's text is pulled out page by page with pdf.js (the reader engine built into Firefox, running here as a local library), and the two texts are aligned by a word-level longest-common-subsequence diff — the same family of algorithm programmers use to compare file versions. Words both documents share pass through plainly; a run that exists only in the first document is shown removed, and a run that exists only in the second is shown added. A changed sentence therefore appears as what it is: the old words struck through, the new words beside them in green. The summary counts are counted, not estimated — pages and words in each file, words added and removed, and the share of the longer text that survived unchanged. Comparison runs page by page as well as overall, so a change on page 9 is reported on page 9.
Text only — the limit, stated plainly
This tool compares text, and only text. It does not detect a moved paragraph's new position as anything but removal-plus-addition, and it is blind to everything that is not words: a swapped photograph, a recoloured chart, a table whose numbers are images, a signature added to a page — none of these register. Scanned documents whose “text” is really a picture of text extract as little or nothing, and the tool will report two different scans as having identical (empty) text; for scans, compare visually or OCR them elsewhere first. Even with born-digital documents, extraction follows each file's internal reading order, which usually matches the page but can jumble multi-column layouts — if a diff looks confused on a complex page, that is extraction order, not a document change. For contracts and policies, treat this as a powerful first pass that finds wording changes fast — and still read any flagged page with your own eyes before acting on it.
Which is A and which is B
Direction matters to the colours: A is the earlier version, B the later. Removals are what A had that B dropped; additions are what B introduced. Swap the files and the same change reads in reverse — the struck-through words become the green ones. If you are checking whether a counterparty's “minor tidy-up” touched the terms, load the version you sent as A and the version that came back as B, and every edit stands out in sequence. Documents of different lengths are handled page-pair by page-pair; pages beyond the shorter document's end compare against emptiness and show as wholly added or removed, which is correct — that content genuinely is new or gone.
Private by design
Like every CalcViso file tool, this one runs entirely in your browser. The file you choose is read by this page, processed on your device, and offered back as a download — it is never uploaded to a server, never stored, and never seen by anyone else. Close the tab and it is gone from memory.
Worked example
Press the sample button: the tool builds two versions of a two-page supplier agreement. Press Compare documents. The summary reports the pair at roughly 96% identical, and on page 1 exactly one sentence breaks into colour — “thirty” struck through in pink, “forty five” in green — because version 2 extended the payment window from thirty days to forty-five. Page 2 is reported identical. That single flagged sentence is the entire difference between the two files, found without reading either of them.
How to use this tool
- Drop the earlier version into the Document A box and the later version into Document B (or create the two samples).
- Press Compare documents and read the summary first: pages, words, and the identical percentage.
- Work through the page-by-page report — identical pages are collapsed to a line; changed pages show the diff in colour.
- Pink struck-through text is what B removed from A; green text is what B added or changed to.
- Open any flagged page in both files yourself before relying on the difference — the diff finds wording; your eyes confirm meaning.
Why there is a word limit
A word-by-word diff holds a comparison grid proportional to both documents' lengths, and asking a phone browser to diff two 300-page manuals word by word would trade your privacy (everything stays on-device) for a frozen tab. The comparison therefore covers the first 2,000 words of each document — ample for letters, agreements, policies and chapters — and says so on the page when a longer document is truncated. For book-length comparisons, compare section by section: split both files with Split PDF into matching page ranges and diff those.