OPUS-CAT
An open-source machine translation engine that runs offline on Windows, with OPUS-MT models and integrations for SDL Trados Studio and memoQ.
OPUS-CAT runs neural machine translation locally on Windows for professional translators who work in computer-assisted translation (CAT) tools. Its offline engine keeps translation processing on the user's computer, so translators can use machine translation for confidential text without sending it to a cloud translation service.
The engine connects to existing translation software. Dedicated plugins support SDL Trados Studio and memoQ, while Wordfast can use it as a custom translation provider. A Chrome browser plugin connects the local engine to Memsource and XTM. Those browser integrations provide access from web-based CAT tools; the translation engine itself runs locally.
Built on Marian NMT, OPUS-CAT uses pretrained models supplied by the OPUS-MT and Tatoeba-Challenge projects. These cover a wide range of language pairs, giving translators a choice of models for their working languages. The focus is machine translation within a professional translation workflow, with model output available through the CAT software where translators already work.
You can also test a downloaded model in the engine's own text translation window before connecting a CAT tool. OPUS-CAT is open source under the MIT license. Its established workflow is on Windows; a cross-platform beta also runs on macOS and Linux, with the same translation functions and a different interface.
