Bottom Line Up Front (BLUF): Text translation on this page runs entirely on your device: the engine executes in the browser, and your text is never sent to the server or paid for with tokens.
The Bergamot engine is compiled to WebAssembly and runs on the CPU — no GPU or video memory is needed. A single language-pair model is about 24 MB, downloaded once and cached by the browser; after that no internet is required.
When there is no direct model for the selected pair, the engine translates through English as a pivot language, so pairs like Russian → German still work.
It translates the text you paste between the supported languages, shows the result in the second field and lets you copy it or download it as TXT. It is not a generative network: it will not write or rewrite text for you.
Quality is good for major languages (around BLEU 30 for English → Russian); for rare languages it is weaker than large paid services.
Your text stays in the browser: no request carrying it is sent to our server. There is no paid API here and no tokens are charged — the translation is free with no usage limit.
Yes. No tokens are charged and the tool has no paid API.
No. The translation runs in your browser and your text never reaches our server.
Only for the first download of the language-pair model. After that it works offline.
No, the engine runs on the CPU and uses ordinary RAM.