OmniDoc Document ExtractionOmniDoc Document Extraction / 2026 guide
How to choose a free document extraction tool in 2026
Document extraction turns a PDF, a scan or an Office file into structured content: headings, paragraphs, tables and images in reading order, as Markdown and JSON. Free tools differ in how much of that structure survives. Check these eight points with one of your own files before you build anything on top of the output.
Try it with your own fileUpdated November 2026
8 things to check in the output
Heading levels
Chapters and sections should come out as heading levels (#, ## in Markdown), not as bold lines, so the outline of the document survives.
Reading order and columns
In two- or three-column pages, lines from different columns must not be mixed; headers, footers and page numbers should not land in the middle of a paragraph.
Tables
Tables should keep their rows and columns, including merged headers, instead of turning into a run of numbers.
Scans and photos
A scan has no text layer. The tool must recognise the text (OCR) and then the layout; check numbers against the original.
Images and formulas
Pictures should be exported as files with their positions marked in the text; formulas should come out unchanged.
Output formats
Markdown suits people and AI assistants; JSON that keeps paragraphs and tables in reading order suits programs, databases and retrieval (RAG).
File types and limits
Check pages per file, monthly limits and the formats you have: PDF, scans, Word, PowerPoint, Excel, EPUB, HTML.
Privacy
Find out who can see your file and whether you need an account. Do not upload confidential documents to a site that does not say.
What the OmniDoc free extraction does on these points
The OmniDoc free document extraction is a web page; you sign in with a free OmniDoc account.
Structure and reading order
Chapters, paragraphs and lists keep the hierarchy of the original; two or three columns of text are joined in reading order.
Tables and images
Tables come out as Markdown tables and as rows and columns in the JSON; pictures are saved in the images/ folder with their positions marked.
What you download
A ZIP file with document.md (the full text as Markdown), document.json (structured data in reading order), the images and a file list.
Limits and privacy
Up to 501 pages per PDF and no monthly cap; free extractions wait in a queue. Only you can see the files you upload and their results.
Which way for which file
| Your file | What to use |
|---|---|
| Research paper (PDF, two columns) | The free extraction: Markdown with heading levels in reading order; check complex formulas against the original. |
| Scanned contract or photo of a page | The free extraction with text recognition (OCR); check numbers and tables against the original. |
| Report full of tables | The free extraction: Markdown tables, and rows and columns in the JSON, ready for a spreadsheet or a script. |
| Word, PowerPoint, Excel, EPUB or HTML | The free extraction reads the document structure directly; no text recognition is needed. |
| Documents for a knowledge base or RAG | Use document.json (paragraphs and tables in reading order) for retrieval and document.md for an AI assistant. |
| Old .doc, .ppt and .xls files | Save them as .docx, .pptx or .xlsx first, then upload them. |
Questions
Is there a free tool to convert PDF to Markdown and JSON in 2026?
Yes. The OmniDoc free document extraction turns a PDF, a scan or an Office file into a ZIP with document.md, document.json, the images and a file list. Up to 501 pages per PDF; sign in with a free OmniDoc account.
How is structured extraction different from copying the text?
Copying gives you plain text. Extraction keeps the headings, paragraphs, tables and images and how they relate, as structured data that a program can use.
Can it extract tables from a PDF?
Yes. Tables come out as Markdown tables and as rows and columns in the JSON. Common merged headers are recognised; check very complex tables against the original.
Does it work for scanned PDFs and photos?
Yes. Scans and images go through text recognition first and are then organised by their layout. Check numbers and tables against the original; results for handwriting depend on the writing.
Can I use the result for RAG or an AI assistant?
Yes. document.md can be pasted into an AI assistant directly; document.json keeps the structure of paragraphs and tables and suits retrieval (RAG).