LibreTranslate provides a free, open-source, and self-hostable machine translation API. It is designed for developers and users who need an alternative to proprietary translation services, offering offline capabilities and ease of setup. The core benefit is a privacy-focused, customizable translation solution.

How It Works

LibreTranslate leverages the Argos Translate library, which utilizes neural machine translation models. This approach allows for offline translation and avoids reliance on external, proprietary APIs. Users can select specific languages to load, reducing memory footprint and startup time. The system supports GPU acceleration via CUDA for improved performance.

Quick Start & Requirements

• Install: pip install libretranslate
• Run: libretranslate [args]
• Prerequisites: Python 3.8+ recommended. CUDA 12.4.1 and nvidia-docker for GPU acceleration.
• Docker: ./run.sh [args] or docker compose -f docker-compose.cuda.yml up -d --build for CUDA.
• Docs: API Docs

Highlighted Details

• Self-hosted and offline-capable machine translation.
• Powered by the Argos Translate library.
• Supports GPU acceleration with CUDA.
• Extensible with various language bindings and integrations (e.g., Discourse plugin, mobile apps).

Maintenance & Community

• Active development with community contributions encouraged.
• Community forum available for support and discussion.
• Roadmap includes language bindings for more languages and improved translations.

Licensing & Compatibility

• License: GNU Affero General Public License v3.0.
• Compatibility: The AGPLv3 license requires that any modifications or derivative works distributed to users must also be made available under the same license. This may impose restrictions on linking with closed-source applications.

Limitations & Caveats

• The AGPLv3 license may restrict commercial use in closed-source environments.
• Performance and translation quality are dependent on the underlying Argos Translate models.
• While offline capable, initial model downloads are required.