Is MarkItDown Online free to use?

Yes. MarkItDown Online is free to use with an optional Pro plan for advanced features.

What file formats are supported?

MarkItDown Online supports PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx), HTML, images (OCR), and audio files.

Is my document data stored?

No. All document processing happens in your browser or via temporary in-memory processing. No uploaded files are stored permanently.

Does MarkItDown Online use Microsoft MarkItDown?

Yes. MarkItDown Online uses the open-source Microsoft MarkItDown library to convert PDF, DOCX, XLSX, PPTX, HTML, images, and audio to Markdown.

Can I convert multiple files at once?

Yes. MarkItDown Online supports batch file conversion. You can upload multiple files and convert them all at once.

Scanned Pdf Ocr Markdown 0820 — MarkItDown Online

PDF conversion now applies OCR to scanned pages and image-based documents

2026-08-20

Starting 2026-08-20, MarkItDown Online applies OCR during PDF conversion whenever a page has no text layer. Previously, scanned contracts, photographed whiteboards, and image-only PDF exports produced empty or near-empty Markdown, because there was no embedded text to extract. Now the converter detects such pages automatically, recognizes their text in 12 languages including English, Chinese, and Japanese, and merges the result into the same Markdown output as regular pages.

How OCR detection works

The converter inspects each page of a PDF before extraction. Pages that carry an embedded text layer go through the normal fast path unchanged. Pages with no text layer — scans, photographs, and documents flattened to images — are routed through OCR, which recognizes the text at up to 300 DPI and groups it into paragraphs and headings by layout. Mixed documents are handled page by page, so a 40-page PDF with three scanned annexes converts in one run. Recognition typically adds 1-2 seconds per scanned page.

What this means for your workflow

Archived paperwork no longer needs a separate OCR step before ingestion: scanned contracts, signed documents, and legacy exports flow through the same batch as born-digital files, and the Markdown preview plus the validation gate catch low-confidence pages before download. Every OCR'd export keeps the standard YAML front matter block, so provenance tracking works unchanged. Feedback: sapsap@qq.com.