Recover an editable Markdown draft and check its reading order
Recover editable notes, documentation or an article draft from a PDF. The converter reconstructs positioned text, proposes headings and lists, and recognizes simple table-like rows. It creates a .md file while keeping the PDF unchanged.
Review the output as a draft. Positioned letters do not always reveal how paragraphs, columns and footnotes relate. The converter can combine neighboring fragments that belong to separate columns, so compare the source and Markdown side by side.
Decide how much structure the source can support
| Your situation | Recommended approach |
|---|---|
| Single-column digital prose | Start with OCR unchecked. Direct text extraction avoids recognizing letters that already exist as text in the PDF. |
| Pages with little or no readable text | Enable local English OCR if the source language and scan quality suit it. Check notices for pages skipped after the device's OCR allowance. |
| Multi-column papers, invoices or complex tables | Plan a manual review of reading order and cells. A result that contains all the words can still associate a number with the wrong row. |
| You need exactly the original appearance | Keep the PDF as the presentation copy. Markdown will not preserve fonts, exact page breaks, diagrams or interactive form controls. |
Recover two sections from the research-notes sample
The sample contains two digital pages headed Browser privacy and Document checks. Its simple layout makes missing or reordered text easy to notice.
- Choose research-notes.pdf and uncheck Use local English OCR when a scanned page has no readable text.
- Select Convert PDF to Markdown and inspect Generated Markdown. The first generated title comes from the filename, so expect research-notes above the document sections.
- Find the Browser privacy and Document checks headings and compare their paragraphs with the corresponding source pages.
- Download Markdown and open the .md file in a Markdown editor. Compare both the raw text and rendered view before reusing the content.
Expected result. The output contains a filename-derived title, inferred section headings and extracted paragraphs in order. It is an editable text file, not a new PDF. The supplied digital-text example does not establish OCR accuracy or complex-table fidelity.
Check the downloaded result
- Compare the first and last paragraph of each section. Check page boundaries where a sentence, footnote or table may continue onto another page.
- Review every number next to its label. Simple table detection uses aligned text gaps, and assumes the first detected row is a Markdown header.
- Inspect hyphenated words and punctuation. Wrapped lines may be joined, while repeated text near page margins can be removed as a presumed header or footer.
- Review the final Links section when present. Supported PDF link annotations are collected there rather than restored to their original inline positions; verify destinations before publishing.
Troubleshooting
- Text from two columns is mixed together
Line reconstruction groups text by its vertical position and then left-to-right position. Separate columns can therefore appear to belong to one line.
Try this: Correct the draft against the source, or extract a smaller section to reduce the review workload. Do not publish reordered financial or scientific text without checking its relationships.
- A table has the wrong header or missing cells
Only conservative patterns of two to six aligned cells across successive lines are treated as tables. Merged cells, irregular spacing and image-based tables do not fit that pattern reliably.
Try this: Rebuild the affected table manually and compare each cell. For a complex table, preserve the original PDF as the reference rather than relying on the generated grid.
- Pictures are missing or links appear at the end
The converter exports text, not embedded pictures. It gathers supported HTTP, HTTPS and mailto link annotations into a separate list and does not reproduce their page placement.
Try this: Extract required images separately and insert them deliberately in your editor. Restore meaningful inline links after checking the destination and source context.
- Scanned text is missing, inaccurate or incomplete
OCR is English-only, runs on pages with insufficient native text, and has a device-dependent page cap. A mixed page with enough existing text may not trigger OCR for its image content.
Try this: Read any skipped-page notice. Process the relevant scans separately and compare names, dates and numbers with the visible page before combining the results.
Limits and compatibility
- Layout limits
- Headings are inferred from text appearance; a PDF's bookmark hierarchy is not used to rebuild Markdown sections. Equations, diagrams, complex reading order and tagged accessibility structure need separate review.
- Processing and export
- Text analysis happens on this device; optional OCR engines load when needed. Copy result requires clipboard permission, while Download Markdown provides the same text as a file.
Your files and form values are processed on this device. Downloaded copies remain wherever you save them. See how to check local processing and our privacy policy for site data and optional services.
Common questions
- Why does the exported document have an extra title?
- The converter adds a top-level Markdown title derived from the source filename. Rename or remove it in your editor if the PDF's own title already supplies the heading you need.
- Can I edit the text directly in the result box?
- The preview is read-only. Copy the result or download the .md file and make corrections in a text or Markdown editor.
Continue your document workflow
Maintained by Shakti Prasad Ram. Found a mismatch? Report the tool, browser and steps without sending a sensitive document. Back to the tool