VisualNovelTranslator
This is a translator plugin for vscode base on Tencent HY-MT1.5 model. You need to download the model on hugging face and use llama.cpp to create a local sever on your PC to use this extension.
Create a local translation server
- Download HY-MT1.5 model on hugging face (or the mirror website, e.g. hf-mirror).
HY-MT1.5-1.8B-Q4_K_M.gguf model is good enough for most cases.
- Download compiled
llama.cpp on github, you can choose any recent version, Windows x64(CPU) version is fast enough and there is no any other dependencies.
- Run the command below in
cmd or powershell
llama-server.exe -c 4096 -m E:\Programming\NN-Framework\HY-MT1.5-1.8B-Q4_K_M.gguf --context_shift -np 1 --port 8080
when you see the text shown in the image below, the server is ready.
The meaning of the parameters:
-m: The path of the model.-c: The size of the context.-np: The number of threads. Only 1 thread is fast enough.--port: The port of the server.--context_shift: Always use this option to allow the model shift the KV cache if it exceeds the context size.
Usage of the plugin
Use the chat window
Press Ctrl + Shift + P to show all the command, and type Open Translation chat,
Click this command and open the chat panel.
Sometimes the model does not work properly, for example, it does not translate the text to the target language. You can click the regenerate button to try again or click the Clear button to clear the translation history.
Translate the selected text
Right click and click the Send to translation chat command. The selected text will be send to the chat panel and translated automatically.
Translate the renpy script
When you edit the rpy translation file (the script file of renpy engine), the text can be translated automatically by press Ctrl + Alt + T
The picture above shows a translation file of renpy. Just put the cursor on the line below the comments, and press Ctrl + Alt + T. The text in the quotation marks on the previous line will be translated and fill into the current line automatically.
