An offline translation server with ultra-low resource consumption and ultra-fast consumption requires only CPU + 1G memory to run without a GPU. The average response time for a single request is 50 ms. Support translation of major languages around the world. The translation quality is comparable to Google Translate. Note that this model focuses on speed and private deployment of multiple devices, so the translation quality is definitely not as good as that of the large model translation. Need high-quality translations suggest using the online big model API. Because the server-side program has no interface, to use help, please open it through a browser and then view it through/help. For authentication, please edit the config.ini file under the application data Model download address Domestic https://ocn4e4onws23.feishu.cn/drive/folder/C3kffkLr8lxdtid5GYicAcFAnTh github https://github.com/xxnuo/MTranServer/releases/tag/models After downloading the models as needed, unzip the zip package for each language into the models folder under the application data. Downloaded the current folder structure diagram of the English-Chinese model: compose.yml models/ ├── enzh │ ├── lex.50.50.enzh.s2t.bin │ ├── model.enzh.intgemm.alphas.bin │ └── vocab.enzh.spm If you download and add multiple models, this is a schematic diagram of the structure of the model folder with Chinese-English and Chinese-English translation: compose.yml models/ ├── enzh │ ├── lex.50.50.enzh.s2t.bin │ ├── model.enzh.intgemm.alphas.bin │ └── vocab.enzh.spm ├── zhen │ ├── lex.50.50.zhen.t2s.bin │ ├── model.zhen.intgemm.alphas.bin │ └── vocab.zhen.spm There is no need to download models that are not available. Download the model according to your needs. Note: For example, the process of translating Chinese into Japanese is to translate Chinese into English first, and then English into Japanese, which means two models, zhen and enja, are needed. By default, the two models of Chinese-English translation and English-Chinese translation have been integrated, and the translation process in other languages is similar. More models https://github.com/mozilla/firefox-translations-models
