Translate a PDF

A clean document if you just want to read it — it always works and always looks tidy. Side by side if you need to check the translation against the original, which is what contracts and study material usually need. Keep the layout if the shape of the page matters, like a form or a certificate.

Because a translation is rarely the same length as the original. German and Vietnamese both tend to run longer than English, and the space on the page does not grow. Where the translated text does not fit, it is set smaller until it does. If a block was tight to begin with, the result can end up noticeably smaller than its neighbours — which is why the clean and side-by-side options exist.

A PDF embeds the fonts it uses, so the tool can only write characters its font contains. A font covering Chinese, Japanese and Korean is tens of megabytes — too much to send to a browser for one translation. Arabic has a different problem: its letters change shape depending on their neighbours, and that shaping cannot be done here.

Both will come later through server-side rendering. In the meantime, translating a Word document has no font limit at all, because Word supplies the fonts.

Then the PDF is a scan or a photo rather than a text document. The page looks like words to you but contains only an image, so there is nothing to extract. Reading one needs OCR, which is not supported yet.

No. The file is opened in your browser, which pulls out the text. Only that text — not the file — is sent to be translated, and the new PDF is built on your device. The text does leave your browser, because something has to translate it. The file, with its images and metadata, does not.