3 ms·

No way to use the offline translation engine standalone with some custom text, or even invoke it from a command line ?

• baobob 4y ago
  There is, it's not even hard to find out how.
  • pabs3 4y ago
    This is the underlying translation engine: https://github.com/marian-nmt/marian-dev https://github.com/marian-nmt/marian-dev And the browser fork of it: https://github.com/browsermt/marian-dev https://github.com/browsermt/marian-dev
    • pluc 4y ago
      I thought it was Bergamot? https://browser.mt/ https://browser.mt/
      • pabs3 4y ago
        Yeah, Bergamot is the fork of the original project (notice the browsermt in the second URL).
• cyber_kinetist 4y ago
  Maybe you can develop it, if you want. The models and the source code for running it are all there, someone just have to wrap it in a standalone binary instead of a WebExtension. - https://github.com/mozilla/firefox-translations https://github.com/mozilla/firefox-translations - https://github.com/mozilla/firefox-translations-models https://github.com/mozilla/firefox-translations-models - https://github.com/mozilla/bergamot-translator https://github.com/mozilla/bergamot-translator Note that most of the heavy NLP code used (MarianMMT, fasttext) were originally written as C++ libraries but were compiled to WebAssembly to make a WebExtension out of it. So if you want to make a command line / GUI desktop app there are two choices: use C++ directly if you want to make an app with minimal dependencies, or use JS/WebAssembly to make a NodeJS / Electron based app (which might be heavier but more accessible for JS devleopers.)