# PDFMathTranslate — Translate PDF Papers Preserving Format

> Translate PDF scientific papers while preserving math formulas, charts, and layout. Supports Google, DeepL, OpenAI, Ollama. CLI, GUI, MCP, Docker, Zotero plugin.

## Install

Copy the content below into your project:

# PDFMathTranslate — Translate PDF Papers Preserving Format

## Quick Use

```bash
pip install pdf2zh

# Translate a PDF (entire document)
pdf2zh input.pdf

# Translate specific pages with DeepL
pdf2zh input.pdf -p 1-10 -s deepl

# Use OpenAI for translation
pdf2zh input.pdf -s openai:gpt-4o

# Use local Ollama model
pdf2zh input.pdf -s ollama:qwen2.5
```

**GUI mode:**
```bash
pdf2zh -g
# Opens a web interface at http://localhost:7860
```

**Docker:**
```bash
docker run -d -p 7860:7860 byaidu/pdf2zh
```

## Intro

PDFMathTranslate (pdf2zh) solves the hardest problem in academic PDF translation: preserving mathematical formulas, charts, tables, and layout while translating text. Most translation tools butcher LaTeX equations and destroy figure layouts — pdf2zh keeps them pixel-perfect.

Featured at EMNLP 2025, it supports 10+ translation backends including Google Translate, DeepL, OpenAI, Anthropic, and local models via Ollama. Integrates with Zotero for one-click translation of your research library. Available as CLI, GUI, Docker, and even an MCP server.

## Details

### Translation Backends

| Backend | Quality | Speed | Cost |
|---------|---------|-------|------|
| **Google Translate** | Good | Fast | Free |
| **DeepL** | Excellent | Fast | Free tier available |
| **OpenAI (GPT-4o)** | Excellent | Moderate | API cost |
| **Anthropic (Claude)** | Excellent | Moderate | API cost |
| **Ollama (local)** | Good | Depends on GPU | Free |
| **Azure Translate** | Good | Fast | Free tier |

### What Gets Preserved

- **Math formulas** — LaTeX equations rendered correctly in output
- **Charts and figures** — untouched, positioned correctly
- **Tables** — structure preserved, text translated
- **References and citations** — formatting maintained
- **Page layout** — columns, margins, headers/footers kept
- **Hyperlinks** — clickable links preserved

### Zotero Integration

Install the Zotero plugin from [GitHub Releases](https://github.com/Byaidu/PDFMathTranslate/releases). Right-click any PDF in your Zotero library → "Translate with pdf2zh". Translated PDF appears alongside the original.

### Output Options

- **Dual-column PDF** — original + translation side by side
- **Translated PDF** — clean translated version only
- **Bilingual mode** — original text with translation below each paragraph

## Frequently Asked Questions

**Q: How accurate is the formula preservation?**
A: pdf2zh uses a dedicated math detection pipeline. LaTeX formulas are extracted, left untranslated, and re-inserted. Accuracy is 95%+ for standard academic papers.

**Q: What languages are supported?**
A: Depends on the translation backend. Google Translate supports 100+ languages. The tool itself is language-agnostic.

**Q: Can I translate a 500-page textbook?**
A: Yes, but use page ranges (`-p 1-50`) for better control. Large documents may need more API quota.

**Q: Does it work offline?**
A: Yes, with Ollama backend. Install a translation-capable model and translate without internet.

## Works With

- Any PDF scientific paper
- Google Translate, DeepL, OpenAI, Anthropic, Ollama
- Zotero reference manager
- Docker, Python 3.12+

## Source & Thanks

- **GitHub**: [Byaidu/PDFMathTranslate](https://github.com/Byaidu/PDFMathTranslate) — 32,600+ stars, AGPL-3.0 License
- **PyPI**: `pdf2zh`
- Featured at EMNLP 2025

---
<!-- ZH -->
# PDFMathTranslate — 保留格式翻译学术 PDF

## 快速使用
```bash
pip install pdf2zh
pdf2zh paper.pdf
```

## 介绍
翻译学术 PDF 同时完美保留数学公式、图表和版式。支持 Google/DeepL/OpenAI/Ollama 等 10+ 翻译后端。EMNLP 2025 入选。32,600+ stars，支持 CLI、GUI、Docker、Zotero 插件。

## 来源与感谢
- **GitHub**: [Byaidu/PDFMathTranslate](https://github.com/Byaidu/PDFMathTranslate) — 32,600+ stars, AGPL-3.0 许可证


---
Source: https://tokrepo.com/en/workflows/4c628f43-c803-45c8-ae39-a4caded80419
Author: TokRepo精选