Translate LaTeX-typeset PDFs to English. All math notation stays as searchable vector text.
This tool is in beta. Feedback welcome.
or browse files
Your PDF is processed on my server and not stored after translation. The first request may take a minute while the server wakes up. Translation speed depends on document length, as each text block is sent sequentially to Google Translate's free API. The server runs on a free hosting tier, so processing is slower than a dedicated machine.
The tool reads PDF text spans directly and classifies each one as either translatable text or mathematical notation based on the embedded font (e.g. CMMI, CMSY, CMEX for math). Text spans are sent to Google Translate with placeholder tokens where math appears. The translated text is then re-rendered into the PDF using CMU Serif and Latin Modern Math, keeping all output as searchable vector text.
Bold, italic, and bold-italic styles are preserved. Table of contents entries, dot leaders, page numbers, and hyperlink annotations are regenerated. Pure-text blocks (without math) are merged into paragraphs before translation for more natural results.
Clone the repo and run with
uv. No configuration needed.