Open-source offline translation library written in Python, Argos Translate targets developers and users needing robust translation capabilities without internet access. It leverages OpenNMT and CTranslate2, offering a flexible solution usable as a Python library, command-line tool, or GUI application. Its primary benefit is extensive language support, achieved through automatic intermediate language pivoting, which expands translatable pairs beyond directly supported models.

How It Works

The library employs OpenNMT for its core translation models, optimized for performance via CTranslate2. Language models are distributed as installable .argosmodel packages managed by the argospm tool. A key feature is its ability to chain translations through intermediate languages (e.g., A -> B -> C) when a direct A -> C model is unavailable, significantly expanding the range of translatable language pairs at a potential minor cost to quality.

Quick Start & Requirements

Installation is straightforward via pip: pip install argostranslate. A GUI is available with pip install argostranslategui. GPU acceleration can be enabled by setting the ARGOS_DEVICE_TYPE environment variable to cuda or auto. Further details and examples are available on the Argos Translate Website and Docs.

Highlighted Details

• Supports over 30 languages directly, with extensive coverage via intermediate language pivoting.
• argospm command-line tool simplifies package management (update, install, list).
• Includes dedicated libraries for HTML (translate-html) and file translation (argos-translate-files).
• Forms the foundation for the LibreTranslate API and web application.

Maintenance & Community

Contributions are welcomed via GitHub issues and pull requests. Support is primarily available through the LibreTranslate Forum and GitHub Issues. The project is sponsored by TestMu AI.

Licensing & Compatibility

Argos Translate is dual-licensed under the permissive MIT License or Creative Commons CC0, allowing for broad compatibility with commercial and closed-source applications.

Limitations & Caveats

Translation quality may degrade when relying on intermediate language pivoting. The graphical user interface is maintained in a separate repository. Custom model training services are offered commercially.