Visually compare two PDF files
A visual diff compares two versions of a PDF as rendered pixels rather than as extracted text. This tool renders the old version (the first file) and the new version (the second file) at the same scale, then produces a PNG image per page in which every changed pixel is drawn red over a faded copy of the page, plus a table showing the percentage of pixels changed on each page. It catches image, layout and font changes that text-based comparison cannot see, and both files stay on your device.
How it works
Comparing PDFs by their text answers only half the question. Text extraction sees the character stream, so it reports what words changed — but a swapped product photo, a nudged logo, a silently substituted font or a table that drifted onto the next page leaves the character stream untouched. Two files can extract identically and print completely differently. A visual diff closes that gap by comparing what the pages actually look like.
The mechanism: both PDFs are rendered page by page with pdf.js, at the same scale so that corresponding points on the page land on the same canvas coordinates. For each page number present in both files, the tool reads the two bitmaps and compares them pixel by pixel. It then composes the output image for that page: a faded, washed-out copy of the page as backdrop, with every differing pixel painted red on top. The result reads at a glance — pale regions are untouched, red regions changed — and each page's ratio of changed pixels to total pixels goes into the summary table, so an 8-page report might show 0.0% on six pages and pinpoint the two that need review. Each difference image can be saved as a PNG for a review email or a change log.
Rendering both files with the same engine at the same scale is what keeps the comparison honest: differences in the output come from the documents, not from two viewers rasterizing differently. The flip side of pixel-level sensitivity is that position matters — text reflow shifts every line below an edit, and the diff faithfully marks all of it as changed, so a one-sentence insertion can redden half a page. That is the correct answer to “does this page render differently?”, but when you want to know which words changed, run the same pair of files through the text-based Compare PDFs tool — the two views complement each other, and reviewing an important document is best done with both.
Worked example: verifying a brochure revision
A print shop returned version 4 of an 8-page brochure (6.1 MB) with the note “minor corrections applied,” and the client wanted to know exactly what had moved since version 3 (5.9 MB). Loading v3 first and v4 second, the diff table read: pages 1, 2, 5, 7 and 8 at 0.0%; page 3 at 2.4% — the red patch sat exactly over one product photo, which had been swapped for a higher-resolution shot; page 4 at 0.3% — a corrected price in a table cell; and page 6 at 11.8%, where a red band spread across the lower two-thirds of the page. That page's edit was a single reworded sentence, but the paragraph reflowed and shifted every following line, which the text-based comparison then confirmed as one changed sentence. Total processing time was around eight seconds, all in the browser.
Frequently asked questions
Which file counts as the old version and which as the new one?
The first file you load is treated as the old version and the second as the new version. Pages are then compared pairwise by page number.
What changes does a visual diff catch that a text comparison misses?
Anything that alters the rendered appearance without changing the extracted words: a swapped image, a shifted logo, a font substitution, changed colors, moved page furniture or reflowed layout. A text comparison sees only the character stream and reports such pages as identical.
What do the red areas in the output images mean?
Red marks every pixel that differs between the two renders of that page, drawn over a faded copy of the page so you can see where the change sits. Unchanged regions stay pale.
Why does one edited sentence light up the whole rest of the page?
Adding or removing words reflows the paragraph, which shifts every line below it by a few pixels, and shifted pixels count as changed. The large red region is technically correct — the page really does render differently from that point down.
Can I visually compare scanned PDFs with no text layer?
Yes, the comparison works on rendered pixels, so it needs no text layer at all. Two separate scans of the same paper will show widespread differences though, because slight misalignment between scanning passes shifts every pixel.
Are the two PDFs uploaded to be compared?
No. Both files are rendered and compared by JavaScript in your own browser, and neither one leaves your device — safe even for confidential contract drafts.