ppt-translator_skill

Overview

This skill translates PowerPoint (.pptx) presentations while preserving fonts, colors, spacing, alignment, and table layouts. It supports multiple LLM providers (OpenAI, Anthropic, Deepseek, Grok, Gemini) and is tuned for language pairs where text expansion or contraction matters (especially CJK ↔ English). Use it to produce translated slides that retain the original visual design without manual reformatting.

How this skill works

The tool extracts slide text into XML, chunks and caches strings, sends them to the selected LLM provider for translation, and reinserts translated text into the original XML while maintaining styling. It applies automatic font scaling for text and tables to prevent overflow when translations expand or contract. Optional intermediate XML files can be kept for debugging and inspection.

When to use it

• Translating single .pptx files between any supported languages
• Batch translating many presentations in a directory
• Translating CJK (Chinese, Japanese, Korean) to/from English where layout changes are likely
• When you must preserve exact slide styling—fonts, colors, alignments, and tables
• When you want to inspect intermediate XML to debug translation or formatting issues

Best practices

• Choose a provider and model that matches your quality and cost needs; test on one slide first
• Use --keep-intermediate to inspect extracted and translated XML if formatting looks off
• Adjust --max-chunk-size to avoid truncation with very long text blocks
• Leverage caching for presentations with repeated headers or phrases to reduce API calls
• For critical layouts, review slides after translation and tweak font scaling if needed

Example use cases

• Translate a corporate presentation from Chinese to English while preserving the brand template
• Batch-translate training decks for global teams (e.g., Japanese → English) with minimal manual reflow
• Localize sales materials into multiple languages while keeping consistent fonts and colors
• Debug layout issues by keeping intermediate XML and iterating on provider settings
• Use a cost-efficient model for large-volume translations (e.g., Gemini) and a higher-capacity model for final polish