How the text is compared
Both pictures are read with Tesseract, compiled to WebAssembly and loaded from this site the first time you use it (about 6 MB for English). The words are normalised — case and punctuation optionally ignored, whitespace always — and aligned with a longest-common-subsequence diff, so an inserted word does not make everything after it look changed.
Each change keeps the boxes of the words it came from, so it can be drawn on the pictures: orange where words exist only in A, blue where they exist only in B.
Prices, dates and other numbers
Turn on “Numbers, prices and dates only” to see just the changes that involve a digit — $49.99 → $39.99, June 12 → June 18 — which is usually what matters on an invoice, a product card or a form.
Where OCR goes wrong
Recognition is good on clean screenshots and scans and weaker on small, stylised, curved or low-contrast text. Every change shows the recogniser's confidence, and anything under 70 is flagged. Treat the list as places to look, not as proof — the boxes on the pictures make that look quick.
Questions people ask
›Can I compare text inside two images?
Yes. Open the Text view and press Compare text. The words in both pictures are recognised in this tab and the differences listed and highlighted.
›Which languages are supported?
English, Spanish, French, German, Italian, Portuguese and Dutch. Each language pack is downloaded once, on request.