EasyNMT

EasyNMT - Easy to use, state-of-the-art Neural Machine Translation is an open-source project which simplifies the usage of state-of-the-art, transformer based Neural Machine Translation (NMT).

Input

Input language:
Target language:
Detected language:	{{(detected_lang in langs) ? langs[detected_lang] : detected_lang}}
Model:	opus-mt

Output

Please wait. A slow $10 CPU machine is translating your text. If it takes more than a second, the respective model must be downloaded which takes up to 60 seconds.

Information

This demo is using the opus-mt model: It provides small models (about 300 MB) for 1200+ different language direction. EasyNMT detects and loads automatically the right translation model. In this demo, we restricted the language selection to about 50 languages.

Notes:

When a language direction is used for the first time, it has to be downloaded, which takes up to 60 seconds. Subsequent calls will be much faster, as the loaded model can be re-used.
Opus-MT sadly doesn't have translation models for all language directions. In that case, have a look at the mBART50 and m2m models, which allow translation between any of 50 and 100 languages.
The input in the demo is currently limited to 100 characters, as a slow CPU server is serving the requests. If you run it locally or with Google Colab, you can translate arbitrary long documents.
See the Google Colab example for testing the other models and for faster translation speed.