A PDF to Markdown converter built for verification
pdfmd turns PDFs into reusable Markdown without hiding how the conversion happened. The browser handles routing and source preview; Cloudflare converts text PDFs, while Moondream 3.1 recognizes fully scanned PDFs.
- Name every boundary
- Local PDF analysis is separate from cloud conversion.
- Evidence over claims
- Unavailable page, coordinate, or confidence evidence stays visibly unavailable.
- Route by document type
- Text PDFs use Cloudflare; scans use Moondream OCR.
Why we built it
Most conversion tools optimize for a fast download. That works until a research paper loses its reading order, a contract drops a footnote, or a table looks correct but attaches values to the wrong headers.
Our PDF to Markdown converter keeps page-level provenance beside the output, making quality visible before the Markdown reaches documentation, search, AI, or knowledge-management workflows.
Local analysis, explicit cloud paths
Text-based PDFs are analyzed in the browser for page structure and preview, then uploaded to Cloudflare's document converter for Markdown output.
Image-only PDFs are rendered into page images in the browser and sent to Moondream 3.1 OCR. The Word to Markdown converter remains entirely on-device, and the image camera clearly labels its cloud upload.
What you can verify
Each sample names its public source and accurate page count, then loads a real PDF through the same conversion path as a file you choose. The result exposes only the page, coordinates, and confidence the converter actually returns.
Editorial responsibility and corrections
Product explanations, guides, benchmark methods, and policy pages are written and maintained by pdfmd's independent operator. Claims are checked against the live conversion routes, billing catalogue, public sample files, and their recorded checksums before publication.
A missing source, factual error, or reproducibility problem can be reported to support@pdfmd.org. Corrections are applied to the page and its structured data together so readers and crawlers receive the same account.
How this gets paid for
Cloud PDF conversion, scanned-page OCR, and image recognition are paid with page points. A text-PDF page costs one point; an OCR or image page costs four. Failed conversions cost zero points.
The browser-local .docx converter remains public and costs zero points. Occasional work can use permanent points, while Pro replenishes a monthly allowance for recurring document workflows.
The paperwork, where you can find it
A paid tool owes you three documents that a product page usually skips: what you are agreeing to, what happens to your data, and when money comes back. Each is written to be read, not to be scrolled past.
- Terms of service
- Points, subscriptions, acceptable use, and the limits of what conversion output is.
- Privacy policy
- The full inventory of what is handled, who receives it, and how long it survives.
- Refund policy
- The 14-day rule for unused points, and what a refund does to an account.
Something that needs a person instead of a policy? Write to support
Where photos and handwriting fit
The browser first checks a PDF's text layer with PDF.js. Text PDFs go to Cloudflare toMarkdown; fully scanned PDFs are rendered page by page in the browser and sent to Moondream 3.1 OCR. Word files still convert in the browser, and the Markdown camera uploads only the compressed images you confirm.
Go deeper by problem
The overview establishes the boundaries. These focused guides break out the verification steps for tables, scans, and knowledge-base imports.
Check the output with a real PDF
Page analysis starts in the browser; cloud PDF conversion requires Google sign-in.