Use LLM to translate Office documents

Translate PowerPoint (pptx) and Word (docx) files using LLM.

Features

• Translate PPTX and DOCX files
• Translation caching for faster repeated translations
• Customizable prompts via llm_prompt.txt
• Configurable via .env file

Setup

1. Apply an API Key

Get your API key from SiliconFlow.

2. Choose a Model

Recommended models: Qwen/Qwen3-8B

Browse available models at SiliconFlow Models.

3. Install Dependencies

pip install openai python-pptx python-docx

4. Configure

Copy .env.example to .env and fill in your settings:

cp .env.example .env

.env Configuration:

Variable	Description	Default
LLM_API_KEY	Your API key	Required
MODEL_NAME	Translation model	Qwen/Qwen3-8B
ENDPOINT	API endpoint	https://api.siliconflow.cn/v1
TEMPERATURE	Response creativity (0.0-2.0)	0.7
ENABLE_THINKING	Enable thinking mode	false

Customize Prompt:

Edit llm_prompt.txt to modify the translation instructions. Use {target_language} as a placeholder for the target language.

Usage

python OfficeDoc_Translator.py <input_file> <target_language>

Examples:

# Translate to Chinese
python OfficeDoc_Translator.py ./input.pptx zh-CN

# Translate to English
python OfficeDoc_Translator.py ./document.docx en-US

# Specify file type explicitly
python OfficeDoc_Translator.py ./file.pptx ja-JP --type ppt

Arguments:

Argument	Description
input_file	Input PPTX or DOCX file
target_language	Target language code (default: zh-CN)
--type	Force file type (ppt or word)
--no-cache	Disable translation cache

Supported Languages:

zh-CN, en-US, ja-JP, ko-KR, fr-FR, de-DE, es-ES, etc.

Cache

Translations are cached in the cache/ directory. Cache is automatically loaded and saved.

Use --no-cache to disable caching.