kikoeru-translate-lrc

A tool that lets you translate various lyric files. Originally created for use with Kikoeru but adjusted to work with any folder structure.

Usage

1. Download the latest release from here and extract it.
   • The correct file is named translate-lrc.zip.
2. Open the .env file in notepad and adjust the settings as necessary.
   • See Configuration for more details on the .env file
3. Save your changes and close the file
4. Place the files you want to translate in the queue folder (or whatever folder you set in the .env file)
5. Run translate.exe
6. Translated files will be placed in the output folder by default

Configuration

To adjust the configuration, you will need to create a .env file in the same directory by creating a copy of the .env.example file.

Common config

• TRANSLATOR - The translation method to use. The available options are:
  • LLM - For any OpenAI API-compatible endpoint. Examples include OpenRouter, llama.cpp, Ollama, LMStudio, etc.
  • DEEPL - For DeepL
  • LIBRE - For LibreTranslate
  • LUNA - For LunaTranslator
• RJ_PATH - The location of the folder that contains all of your lyric/subtitle files that need to be translated. This path will be searched recursively. Defaults to ./queue in the same directory if not provided.
• BAK_PATH - The location of the folder where you want a backup of your lyric/subtitle files to be stored. Defaults to ./backup in the same directory if not provided.
• OUT_PATH - The location of the folder where you want your translated lyric/subtitle files to be stored. The folder structure of the files will be the same structure as the RJ_PATH folder. Defaults to ./output in the same directory if not provided.

LLM-specific config

These configuration items only apply if your TRANSLATOR is set to LLM:

• LLM_ENDPOINT - Required. The Chat Completions URL. Must be OpenAI API-compatible.
• LLM_API_KEY - The API Key or other Bearer auth token to used for authenticaing with the LLM service. Can be removed if you are running an LLM locally.
• LLM_MODEL - The model to use for the translation. If this is not included, it will used the already-loaded model (local-only).

You can also override the parameters used for the LLM:

• LLM_PARAM_PROMPT - The system prompt to use for the translation. To improve accuracy, your system prompt should include the desired source and target language.
• LLM_PARAM_TEMPERATURE - Temperature
• LLM_PARAM_TOP_P - Top-P
• LLM_PARAM_TOP_K - Top-K
• LLM_PARAM_REPETITION_PENALTY - Repetition Penalty

DeepL-specific config

These configuration items only apply if your TRANSLATOR is set to DEEPL:

• DEEPL_API_KEY - Required. The DeepL API Key. Requires a DeepL account, but creating your own key is free. You can find more information here.
• DEEPL_SOURCE_LANG - The source language code you want to translate from. If it is unknown or you would prefer to have DeepL auto-detect it, you can delete this entirely. See this article for a list of supported language codes.
• DEEPL_TARGET_LANG - The target language code you want to translate to. Defaults to en-US if not provided. See this article for a list of supported language codes.

LibreTranslate-specific config

These configuration items only apply if your TRANSLATOR is set to LIBRE:

• LIBRE_ENDPOINT - The endpoint which is hosting your LibreTranslate instance.
• LIBRE_SOURCE_LANG - The source language code you want to translate from. If it is unknown or you would prefer to have it auto-detected, you can delete this entirely. See this page for a list of supported language codes.
• LIBRE_TARGET_LANG - The target language code you want to translate to. Defaults to en if not provided. See this page for a list of supported language codes.

LunaTranslator-specific config

These configuration items only apply if your TRANSLATOR is set to LUNA:

• LUNA_ENDPOINT - The endpoint which is hosting your LibreTranslate instance.
• LUNA_TRANSLATOR - The name of the translator to use within LunaTranslator. You must configure the translators yourself within LunaTranslator. You can visit the /api/list/translator endpoint of your server to get a list of translator names.

Running from source

This project requires deno to run.

1. deno install
2. Copy .env.example to .env
3. Edit .env and adjust settings as necessary.
   • See Configuration for more details on the .env file
4. Place your files/folders with untranslated subtitles in your RJ_PATH (queue) folder
5. deno task translate

Tasks

• deno task translate - The main task. Translates lyrics/subtitles as specified in this README.
• deno task deepl-tlfix - Useful for re-translating or fixing a cache file that may have been translated incorrectly by DeepL. For example, if automatic language detection is enabled, but DeepL selected the wrong language. To use:
  1. Copy the bad data/tlcache.json to data/tlfix.json
  2. Strip the file so that only the entries you want to retranslate are there
  3. Replace detectedSourceLang with the correct language code
  4. Run the script. The translated file will be saved to data/tlfix-translated.json