how conversion works
Structure first. Fidelity honestly.
anydoctomd uses format-aware browser parsers instead of treating every file as plain text. The output is designed to be useful and inspectable, not to pretend that a visual document and Markdown are identical.
Office and OpenDocument files
Modern DOCX, PPTX, and XLSX files are ZIP packages containing XML parts. The converter reads document paragraphs, styles, relationships, slide sections, worksheets, shared strings, formulas, and table cells. ODT, ODS, and ODP use their XML content and format-specific structures.
PDF and OCR
Text PDFs are read through their text layer and grouped into page sections using coordinates. When a PDF has no extractable text, the browser can rasterize pages and use OCR. Image OCR follows the same browser-first approach.
HTML, EPUB, CSV, RTF, and text
HTML extraction removes common interactive chrome before collecting headings, paragraphs, lists, quotes, code, and tables. EPUB chapters follow the package reading order. CSV parsing respects quoted fields and common delimiters. RTF and plain text use lightweight state-aware normalization.
Known limitations
- • Visual positioning, animations, floating objects, charts, and embedded media are not fully reproduced.
- • OCR can misread low-resolution scans, handwriting, columns, names, and numbers.
- • URL conversion depends on public access, CORS, and the HTML returned by the target site.
- • AI-ready means cleaned and structured; it does not mean summarized, fact-checked, or error-free.
references
Technical sources.
CommonMark specification ↗
The Markdown syntax model used as a reference for normalized output.
PDF.js documentation ↗
The browser PDF text and page rendering library used for PDF paths.
Tesseract.js ↗
The browser OCR library used for image and scanned-PDF recognition.
W3C EPUB 3.3 ↗
The package, manifest, spine, and content model referenced for EPUB ordering.
Microsoft Open XML overview ↗
Background on the package structure used by modern Office documents.