Hysen Labs
CLI tool
PDFMathTranslate/PDFMathTranslate avatar
PDFMathTranslate

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

36,173 stars3,229 forksPythonAGPL-3.0
01
DEEP OPEN-SOURCE ANALYSIS

PDFMathTranslate: translating papers while keeping the layout

An AI based PDF translation tool that preserves formulas, charts, tables of contents, and annotations, with Google, DeepL, Ollama, and OpenAI among its translation backends.

02
DEEP OPEN-SOURCE ANALYSIS

What the tool does

PDFMathTranslate is a PDF scientific paper translation tool that preserves formatting. The README says it keeps formulas, charts, table of contents, and annotations intact, supports multiple languages, and can be driven as a command line tool, an interactive user interface, or through Docker. The project appears as an EMNLP 2025 demo. Python is the primary language, the license is AGPL-3.0, and the homepage is pdf2zh.com.

03
DEEP OPEN-SOURCE ANALYSIS

Running a translation

Executing the translation command produces two files in the current working directory: a translated document and a bilingual document. Google is the default translation service, with more services available. The repository description names Google, DeepL, Ollama, and OpenAI among the supported backends.

04
DEEP OPEN-SOURCE ANALYSIS

Advanced options

Advanced options cover local files, online files, entering the GUI, partial document translation, source and target language settings, translation service selection, multi threading, and an output directory. The README points to an advanced usage document for the full list of each option, so the main file stays short.

05
DEEP OPEN-SOURCE ANALYSIS

Deployment and regional quirks

Version 2.0 moved to a new repository and the official release has been published. Public free services online are recommended for people who do not want to install anything, including an immersive translate option and a demo hosted on ModelScope. Docker images exist, with a GitHub Container Registry alternative. The README also warns that users in specific regions may hit network difficulties loading the AI model, documents an environment variable workaround, and covers other issues in an FAQ. A Windows tip suggests installing a runtime if a downloaded file will not open.

07
DEEP OPEN-SOURCE ANALYSIS

Official sources

08
Community notes

Community notes