PDFMathTranslate/PDFMathTranslate

Translates scientific PDFs while preserving layout — equations, tables, figures, and column structure all stay where they belong. That sounds trivial until you have tried running an arXiv paper through a generic translation service and watched the LaTeX blow up into noise. The tool supports Google Translate, DeepL, Ollama, and OpenAI backends, outputs bilingual side-by-side PDFs, and ships with CLI, GUI, Docker, and a Zotero plugin. That last one is the buried lede: researchers who live in Zotero can translate directly from their reference manager without a separate workflow step. The breadth of translation backends means you can keep costs near zero with a local Ollama model or flip to DeepL for higher accuracy when it matters. Reservation: quality on heavily notation-dense proofs depends entirely on the backend model you choose. -> Best for: technical PM or ML researcher who regularly reads non-English academic literature