PDFMathTranslate
[EMNLP 2025 Demo] PDF scientific paper translation with preserved formats - 基于 AI 完整保留排版的 PDF 文档全文双语翻译,支持 Google/DeepL/Ollama/OpenAI 等服务
What is PDFMathTranslate?
[EMNLP 2025 Demo] PDF scientific paper translation with preserved formats - 基于 AI 完整保留排版的 PDF 文档全文双语翻译,支持 Google/DeepL/Ollama/OpenAI 等服务,提供 CLI/GUI/MCP/Docker/Zotero
Overview of PDFMathTranslate
What it is
nPDFMathTranslate is a tool for translating scientific PDF documents while preserving the original layout, including formulas, charts, table of contents, and annotations. It is presented as an EMNLP 2025 Demo and supports multiple translation services such as Google, DeepL, Ollama, and OpenAI. The tool provides several interfaces: command-line tool (CLI), graphical user interface (GUI), Docker, and Zotero integration. n n
Key features
n
- Preserves formulas, charts, table of contents, and annotations during translation. n
- Supports multiple languages and diverse translation services (Google, DeepL, Ollama, OpenAI, MiniMax, etc.). n
- Provides command-line tool, interactive user interface, Docker, and Zotero integration. n
- Includes experimental OCR support with paragraph regrouping and adaptive typesetting. n
- Offers an
--ultrafastCLI mode for native-text PDFs that uses heuristic layout without ONNX layout and OCR. n - Version 2.0 has moved to a new repository (PDFMathTranslate/PDFMathTranslate-next). n n
Use cases
n
- Translating scientific research papers from one language to another while keeping the PDF formatting intact. n
- Bilingual reading of academic documents where the translated text is overlaid on the original. n
- Batch processing of PDF files in a local or Docker environment for research or archival purposes. n n
Getting started
nTo install using Python with uv: n n# Requires Python 3.11+ and uv >=1.25.2 npip install -U pdf2zh n nFor other installation methods (CLI, GUI, Docker), see the repository README. n n
License and pricing
nThe project is licensed under AGPL-3.0 and is open source. No pricing information is stated in the README or metadata.
Based on the project's public README and repository metadata.