Most tools that translate a PDF treat it as a bag of loose text, and the result is a mess: equations break, two-column flow collapses, and tables turn to soup. BabelDOC's bet is that translation and layout are two separate problems. It first parses a paper into a structured model — text blocks, figures, tables, formulas — translates only the language, then re-renders a new PDF that mirrors the original page geometry.
What Sets It Apart
- Parse-then-render pipeline: layout fidelity comes from reconstructing the document, not patching strings in place, so a translated page still reads like the paper it came from.
- Bring-your-own LLM: any OpenAI-compatible endpoint works, so you control cost, quality, and privacy rather than being locked to one provider.
- Bilingual or monolingual output: side-by-side originals make it a study aid, not just a translator — useful when you don't fully trust the model on technical terms.
- CSV glossaries: pin domain terminology across a whole document, which matters more in scientific text than raw fluency.
Who It's For
Great fit if you read English research papers and want a Chinese (or bilingual) copy that keeps figures, tables, and equations where they belong — especially paired with its Zotero and Immersive Translate integrations. Look elsewhere if you need non English-to-Chinese pairs (other directions are largely untested), heavily scanned or image-only PDFs, or pixel-perfect output: it skips very large pages and can misparse author and reference sections.
