Argos Translate

Argos Translate runs neural machine translation locally, so you can translate text without sending it to an online service. It's for people who need offline translation and developers who want to add it to their own software. The Python library and desktop interface use the same translation engine, and the project is open source under the MIT license.

Language models come as downloadable .argosmodel packages. Once the models you need are installed, translation works offline. Supported languages include English, Spanish, French, Chinese, Arabic, Japanese and Ukrainian, among others. You can also train and package custom models with the project's training scripts.

Argos Translate can route a translation through an intermediate language when you don't have a direct model for a language pair. For example, Spanish-to-English and English-to-French models let it translate Spanish into French. That extends the language pairs available from your installed models, though the extra step can reduce translation quality.

Developers can use the Python library or command-line interface. Argos Translate uses OpenNMT for translation and supports CUDA GPU acceleration through CTranslate2. Related libraries add HTML translation through translate-html and document translation through argos-translate-files.

LibreTranslate builds a web app and API on this engine, including support for HTML and files. The hosted demo sends requests to an online service; Argos Translate's local library and desktop app perform translation on your own machine.