Free document translation and structured extraction for whole PDF, Word, PPT and Excel files

If you have ever opened a long PDF in a language you do not read, you know the routine: copy a paragraph, paste it into a translator, copy the result back, repeat. Tables break, headings lose their place, and the original layout is gone.

OmniDoc does the whole file in one go, and both the translation and the structured extraction are free.

Free document translation

Open the free translator, sign in (registration is free), drop in a file, pick the target language and download the result: https://omnidoc.top/free/translate/

• Whole files, not snippets. PDF, Word, PowerPoint, Excel, CSV, EPUB, Markdown, TXT, HTML, SRT/ASS subtitles, and JPG/PNG images.
• Layout stays where it was. Headings, paragraphs, tables and images keep their positions as far as possible. Table cells are translated one by one, so rows and columns survive.
• Bilingual and translation-only. For PDF and LaTeX you get both: a side-by-side version for checking against the original, and a clean translation for reading. PowerPoint comes back as translation-only.
• Source language is detected for you. Target languages include Chinese, English, Japanese, Korean, French, German, Spanish and more, 30+ in total.
• Limits: up to 501 pages per file, two files processed at once, the rest wait in a queue. Your files and results are visible only to you.

It is machine translation, so for contracts and papers please check numbers, terms and citations against the original.

Free structured extraction

Scans and PDFs are awkward to reuse. The free extractor turns PDFs, scans, Word, PowerPoint and Excel files into Markdown and JSON: https://omnidoc.top/free/extract/

PDFs, scanned PDFs and images go through layout analysis, OCR and table recognition. Word, PowerPoint, Excel, EPUB and HTML are parsed directly. You download one ZIP:

document.md      the full text as Markdown, headings and tables kept
document.json    structured data in reading order
images/          pictures, with their positions marked in the Markdown and JSON
manifest.json    file list and basic information

Tables come out as Markdown tables and structured data instead of a run of plain text, which makes the result usable for RAG pipelines, note-taking, or feeding to an LLM. Same limits as translation: up to 501 pages per PDF, two files at a time.

And a desktop app that works on the files themselves

OmniDoc also ships as a free-to-download desktop app for Windows, Linux and macOS (Intel). You can edit documents, spreadsheets, presentations and PDFs. In a PDF you click into a paragraph and type, and the layout you see while typing is the layout that gets saved. Export to Word, Excel and PowerPoint is built in.

There is an MCP interface too, so Claude, Codex, Cursor and other AI tools can read, edit and export documents directly, and an open document updates in place as the AI changes it. Setup is described in the MCP repo: https://github.com/CBIhalsen/omnidoc-mcp

• Installers and checksums: https://github.com/CBIhalsen/omnidoc/releases
• Website: https://omnidoc.top

In the team's words

Free translation, so language barriers get smaller. Free for students, so good tools get closer to learners. We are aiming for the strongest translation and structured extraction, and for the world's first dynamic-content document. OmniDoc: make documents productive.

OmniDoc is proprietary software, free to download and use. Online features (accounts, translation, extraction, AI) are governed by the terms at https://omnidoc.top/terms.