PPT Translator - Tradução com Formatação Intacta
tristan-mcinnis/PPT-Translator-Formatting-Intact-with-LLMs
Traduz PowerPoint preservando formatação com suporte multi-provider
Ver no GitHubFerramenta Python que traduz apresentações PowerPoint mantendo toda a formatação original (fontes, cores, layouts, tabelas). Suporta múltiplos provedores de LLM (DeepSeek, OpenAI, Anthropic, Grok, Gemini) com processamento em lote e cache inteligente para otimizar custos.
Instalação
npx convez-ai tristan-mcinnis/PPT-Translator-Formatting-Intact-with-LLMs1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Traduzir decks corporativos para múltiplos idiomas sem perder design
- Processar lotes de apresentações em diferentes idiomas automaticamente
- Localizar conteúdo visual mantendo consistência de marca
Veredito

Solução prática e bem-implementada para automatizar tradução de PowerPoint. Destacam-se o suporte multi-provider, cache de strings e processamento em lote; ideal para equipes que precisam localizar apresentações em escala sem redesenho manual.
README
A tradução para o português deste README ainda está na fila.
PPT Translator
Convert your PowerPoint presentations to beautifully translated documents while preserving formatting
Clean, fast, and reliable PowerPoint translation with multi-provider support and formatting preservation
✨ Features • 🚀 Quick Start • 📖 Usage • 🧪 Testing • 🤝 Contributing
✨ Features
• ⚡ Lightning Fast: Sub-2 second translation for most presentations • 🔄 Multi-Provider Support: Switch between DeepSeek, OpenAI, Anthropic, Grok, and Gemini with a simple CLI flag • 🎨 Rich Formatting: Preserves fonts, colors, spacing, tables, and alignment after translation • 🔗 Smart Caching: Avoids duplicate API calls for repeated strings • 📦 Batch Processing: Convert entire directories of presentations at once • 🛡️ Robust Processing: Handles all PowerPoint content types with graceful fallbacks
🆕 Recent Updates
- Gemini support: Added Google Gemini as a translation provider
- Updated model defaults: Now using latest model versions (GPT-5.2, Claude Sonnet 4.5, Grok 4.1)
- Security fix: Replaced unsafe
eval()with safe lookup functions
📦 Requirements
- Python 3.10+
- macOS (primary target), Linux, or Windows
- Provider API keys stored in environment variables (see below)
Install dependencies:
python3 -m venv .venv
source .venv/bin/activate # macOS/Linux
# .venv\Scripts\activate # Windows PowerShell
pip install -r requirements.txt
🔐 Configuration
Copy example.env to .env and fill in the API keys for the providers you plan to use. All keys are optional – only populate the providers you intend to call.
cp example.env .env
Environment variables of interest:
| Provider | Required variable | Optional variables | Default model |
|---|---|---|---|
| DeepSeek | DEEPSEEK_API_KEY | DEEPSEEK_API_BASE | deepseek-chat |
| OpenAI | OPENAI_API_KEY | OPENAI_ORG | gpt-5.2-2025-12-11 |
| Anthropic | ANTHROPIC_API_KEY | — | claude-sonnet-4-5-20250514 |
| Grok | GROK_API_KEY | GROK_API_BASE | grok-4.1-fast |
| Gemini | GEMINI_API_KEY | — | gemini-3-flash-preview |
📝 The CLI reads your
.envfile automatically when run from a shell session that has the variables exported. On macOS you can add the exports to~/.zshrcor usedirenvfor project-specific secrets.
🚀 Usage
Run the CLI with the path to a single presentation or a directory tree:
python main.py /path/to/decks \
--provider openai \
--model gpt-5-mini \
--source-lang zh \
--target-lang en \
--max-workers 4
Common options:
--provider {deepseek,openai,anthropic,grok,gemini}– choose the model provider.--model MODEL_NAME– override the default model for that provider (e.g.gpt-5-nano).--source-lang/--target-lang– ISO language codes.--max-chunk-size– character limit per translation request (default: 1000).--max-workers– number of threads used when scanning slides (default: 4).--keep-intermediate– keep intermediate XML files for inspection/debugging.
The tool will generate:
{deck}_original.xml– source deck contents.{deck}_translated.xml– translated content.{deck}_translated.pptx– rebuilt presentation with translated text and formatting intact.
🧪 Testing
Run unit tests with Pytest:
pytest
The test suite focuses on translation chunking/caching and CLI utilities to ensure the core pipeline stays reliable as providers evolve.
🛠️ Project Structure
.
├── .claude/skills/ppt-translator/ # Self-contained Agent Skill
│ ├── SKILL.md # Skill instructions for Claude
│ ├── LICENSE.txt
│ └── scripts/ # Complete translation toolkit
├── ppt_translator/
│ ├── cli.py # CLI parsing and orchestration
│ ├── pipeline.py # PPT extraction, translation, regeneration
│ ├── providers/ # DeepSeek, OpenAI, Anthropic, Grok, Gemini adapters
│ ├── translation.py # Chunking + caching translation service
│ └── utils.py # Filesystem helpers
├── tests/ # Pytest suite
├── example.env # Environment variable template
├── requirements.txt
└── main.py # Entry point (delegates to CLI)
🤖 Agent Skill
This project includes a self-contained Agent Skill in .claude/skills/ppt-translator/.
The skill directory contains everything needed to use this translator in any agent workflow:
SKILL.md- Instructions for Claude on how to use the toolscripts/- Complete copy of all translation scripts and dependencies
To use in another project: Copy the .claude/skills/ppt-translator/ directory into your project's .claude/skills/ folder. The skill is fully self-contained with all necessary scripts included.
🤝 Contributing
Pull requests and issues are welcome. Please run pytest before submitting changes and document any new providers or configuration steps in the README.
📄 License
This project remains under the MIT License. See LICENSE for details.