Compare two PDFs page by page
Choose two PDFs and the page numbers to compare. Pages are paired by number; adjust the threshold to highlight differences between their rendered pixels.
Runs locally in your browser ·
Choose two PDFs and the page numbers to compare. Pages are paired by number; adjust the threshold to highlight differences between their rendered pixels.
Runs locally in your browserChoose two PDFs and press Compare PDFs: the two files are rendered in the browser at 1.2× and the renderings of each page number are compared pixel by pixel. Every page gets a card with a badge, the measured difference and a map in which changed pixels are red, and Download difference images packs the same maps into a ZIP, one page-N-difference.png per compared page.
A three-page pair took about 0.6 s in this browser. Two identical files answered 0.00% on all three pages; one extra line of text on page 2 measured 0.38% at the default threshold of 18 and 0.44% at 0; a second file with two pages reported the third page as missing. Twenty pages are compared per run, so a 22-page pair with an empty Pages field stops with the limit message instead of quietly comparing the first twenty.
Each card carries the page number, a badge (Identical, Changed or Page missing), the measured share of pixels and the map. On the map the pixels the two renderings do not share are painted red; the rest is a light grey built from the average brightness of the two pages, so the shape of the document stays visible under the marks. It is a map, not a side-by-side view and not a draggable overlay.
Two identical files measure 0.00% on every page and stay grey. One added line of text measured 0.38% of the page at the default threshold, so the number is the share of the page above the threshold, not a severity score. The threshold decides where the line between same and changed is drawn: at 255 no pixel can pass it, and the report then says 0.00% while the files still differ.
Pages are paired by number, not by position: if one file is shorter, the pages that exist only in the longer file come back as Page missing with 100.00%, and the map shows the existing page against a blank canvas. Nothing is padded, inserted or guessed.
Each page is rendered at its own size, so two documents with different page sizes measure about 10.8% on every page that exists, even when they carry the same text. Resize or crop the files first if you need to compare the content alone.
Only the rendering is compared: no text layer, no metadata, no embedded files and no logical structure. A change that leaves the rendering identical (corrected metadata, an added bookmark, a swapped font that looks the same) is not reported, and a change that exists only in the rendering, such as a different zoom at export time, is.
pdf.js renders the pages and JSZip packs the maps, both inside the browser: the two files are not uploaded to VoriTools or anywhere else and are closed when the run ends. Damaged, empty or renamed files are refused with a written message, and encrypted or password-protected PDFs cannot be opened at all.