This page is for PDFs with no text you can copy: a letter from your landlord scanned at the office, a certificate with a stamp, a contract someone photographed on their phone and saved as PDF. There is nothing to select in an image, so you can't paste it into a translator bit by bit. Here you upload the file as it is and get the translation back as a Word document.
Every page without a text layer is recognised first, the way an OCR program would do it, and then translated. You don't have to mark anything, the detection is automatic. A PDF can have up to 30 scanned pages and weigh up to 95 MB; in total it can hold up to 250,000 characters, about 100 pages of dense text. If your scan is longer, split it before uploading. A PDF that mixes text pages and scanned pages works as well, and the pages stay in order. The source language is detected for you, and you pick one of 45 target languages.
You get a .docx file that opens in Word or Google Docs, with the text, headings, paragraphs, lists and tables, which are recognised as real tables rather than loose lines. What doesn't carry over: images, fonts, columns and the exact position of things on the page. Stamps and signatures are not copied as images either. If you need a PDF, save one from Word.
For a good result, scan the page straight, in good light and without shadows. A blurry or tilted photo can lose words, so check names, dates and amounts against the original. In our tests a 3-page scan took 44 seconds; for several pages, expect about a minute.
The file is processed on our server and deleted from there after 24 hours. The translation stays in the library in this browser. It is done by AI: a working translation to understand the letter and reply to it, not a certified translation for a court, a consulate or an immigration office. To keep bots out, there is a cap of 60 AI tasks per hour from one connection.
