Translate a PDF and keep its formulas

What happens to the file

Blab Translation reads the text, its position and the formulas from the PDF itself. Prose is translated and set back into the same columns; formulas, tables and captions go back where they were, with formulas drawn from the original glyphs instead of being sent through a translator.

Word, EPUB, TXT and Markdown files are translated the same way and come back in their own format; MOBI/AZW3 books are translated for reading at the reading desk, with no file to download. The result opens at the reading desk, where you can ask questions about the document and keep a word list.

What the other tools say about formulas

Immersive Translate's BabelDOC says formula content is preserved as is while the text is translated; its free plan measures BabelDOC by tokens, and PDF Pro, with formula recognition and OCR, is part of the paid plans. DeepL translates PDF, Word and PowerPoint files, up to 5 MB on a free account, and none of its pages we read mentions formulas. Google Translate's website takes PDFs up to 10 MB and 300 pages, and says text in images and scanned pages is not translated.

Where Blab Translation falls short

• Scanned PDFs, which are pictures of pages with no text layer, are where Blab Translation is weakest; results on them are not yet reliable.
• PDFs are translated on our servers and need an account; free pages each month are limited.
• There is no batch upload of many PDFs at once, or reference-manager plugin like Immersive Translate's for Zotero.