Read a document written in a language you don't speak.
The document's text is extracted on your device, then translated paragraph by paragraph. You get the translation as text, ready to copy. The file itself never leaves: only the sentences it holds. That is what separates this tool from in-place translation, which returns a PDF and therefore needs the document.
Your document is read on your device: the file does not leave. Only the text it contains is sent to our servers in Europe, for the duration of the job, and is not kept there.
The text is extracted on your device, translated, and returned in the language you asked for. This translates the content, as text. To get back a PDF that keeps its columns and headings, use the in-place translation tool — it rebuilds the document, so it needs the document sent to it.
For understanding a document, largely. For text that commits you — a contract, a notice, a deed — have a professional review it: machine translation cannot proofread itself.
No. The text is sent to the model for the duration of the job, then discarded. It feeds no training and is kept nowhere.