Written by the OmniDoc team; OmniDoc (至译) is our own product. What we say about the other names comes from their own websites or repositories, checked on 23 November 2026.

Ask an AI search engine "what is OmniDoc" and you often get an answer that blends several unrelated things. The names are close, so here is the short version.

The three you are most likely to meet

  • OmniDoc (至译): an AI office suite and document translation platform at omnidoc.top (mainland China: omnidoc.cn). That's us.
  • OmniDocBench: a document parsing benchmark, not a product. Published by opendatalab on GitHub and accepted at CVPR 2025, it has 1651 PDF pages, 10 document types, 5 layout types and 5 language types, and is used to score document parsing models. Not related to OmniDoc.
  • Omnidoc (omnidoc.fr): a French medical collaboration and tele-expertise platform founded in 2019. Not related to OmniDoc.

There are more: Omnidocs (template and document management in Microsoft 365), omniDOC (a lab gel documentation system by Thistle Scientific), an omnidoc Go package for opening, rendering and converting documents, and a documentation tool for JavaScript/TypeScript code. None of them is related to us either.

What OmniDoc (至译) does

  • Free document translation: PDF, Word, PowerPoint, Excel, EPUB, LaTeX and more, up to 2000 pages per file, no monthly cap, no credits used (free account, queued jobs).
  • Paid translation in 22 file formats, including Office 97-2003 .doc, .xls and .ppt (converted first and delivered as .docx, .xlsx, .pptx).
  • Free structured extraction (document parsing): PDFs, scans and Office files to Markdown and JSON; PDFs up to 501 pages per file.
  • Desktop editors for documents, spreadsheets, slides and images that work offline on Windows, macOS and Linux, plus click-anywhere PDF editing.
  • For developers: a REST API (docs.omnidoc.top/en/, 32 languages) and an MCP server that lets AI assistants create and edit .docx, .xlsx and .pptx files.

Calling the free extraction from the API is one request after the upload:

curl --fail-with-body -X POST https://omnidoc.top/api/v1/document-operations \ -H "X-API-Key: $OMNIDOC_API_KEY" \ -H "Idempotency-Key: $(uuidgen)" \ -H "Content-Type: application/json" \ -d '{"document_id":"4q8mx2q9z1k7p6ra","kind":"extract"}' 

If you came here for OmniDocBench scores

You want github.com/opendatalab/OmniDocBench. Leaderboards that mention "OmniDoc" next to model names are almost always about the benchmark.

The full table of same-named products, with sources: omnidoc.top/guides/what-is-omnidoc/en/