funstory-ai/BabelDOC

BabelDOC – PDF scientific‑paper translation with bilingual output

What it is – BabelDOC is a Python library / CLI that translates PDF documents (especially scientific papers) from a source language into a target language and produces a bilingual PDF (original + translation side‑by‑side or alternating pages). It drives the translation with an LLM‑based service (OpenAI‑compatible models such as gpt‑4o‑mini, glm‑4‑flash, deepseek‑chat, etc.) and includes a rich set of PDF‑pre‑processing options to keep layout, formulas, tables and fonts intact.

Key capabilities

• Command‑line interface (babeldoc) and a Python API for programmatic use.
• Supports English → Chinese as the primary pair; basic English‑to‑English support is also available.
• Bilingual PDF generation – original and translated pages can be placed side‑by‑side, alternated, or output as separate monolingual files.
• Fine‑grained PDF handling: page selection, line‑splitting, OCR work‑arounds, formula detection, table translation, font overrides, watermark control, and compatibility tweaks for problematic PDFs.
• Glossary injection – CSV glossaries can be loaded so that domain‑specific terms are forced in the LLM prompt.
• Offline assets – a single zip can bundle required fonts and models for air‑gapped deployments.
• Rate‑limit and parallelism – QPS throttling and configurable worker pools.
• Self‑deployment – can be run locally (via uv or a virtual environment) or through the hosted beta service Immersive Translate – BabelDOC (free quota of 1 000 pages/month).

Typical workflow

# install the tool (requires uv and Python 3.12)
uv tool install --python 3.12 BabelDOC

# translate one or more PDFs using an OpenAI‑compatible model
babeldoc \
  --openai \
  --openai-model "gpt-4o-mini" \
  --openai-base-url "https://api.openai.com/v1" \
  --openai-api-key "YOUR_KEY" \
  --files paper.pdf

Options let you change source/target languages (--lang-in, --lang-out), select pages (--pages), enable OCR work‑arounds, control watermarking, supply glossaries (--glossary-files), and more.

Installation

• Recommended: use uv (uv tool install --python 3.12 BabelDOC).
• Alternatively, clone the repo and run uv run babeldoc ….
• The package is published on PyPI (pip install BabelDOC works but the README suggests uv for best experience).

Where it fits

• AI‑driven translation: the heavy lifting is done by LLMs, not traditional MT engines.
• Document‑processing pipeline: combines layout analysis, optional OCR, and LLM prompting in a single end‑to‑end tool.
• Embeddable: can be called from other Python programs or integrated into reference‑manager plugins (e.g., Zotero via immersive-translate/zotero-immersivetranslate or guaguastandup/zotero-pdf2zh).

Status & community

• Beta online service available; free tier provides 1 000 pages/month.
• Open‑source, MIT‑style license (see LICENSE).
• Contributions are welcome; a hiring page is linked in the README.

Useful links

• Online demo: https://app.immersivetranslate.com/babel-doc/
• Self‑hosted backend: https://github.com/PDFMathTranslate-next/PDFMathTranslate-next
• Documentation: https://funstory-ai.github.io/BabelDOC/supported_languages/
• Telegram community: badge in the header.

---

All information above is taken directly from the repository’s README; no additional features have been inferred.