Which part of turning documents into AI-ready Markdown needs the most work?

DocuPrune is launching today, and I would love feedback on the document workflow itself.

A quick way to try it without sharing your own files:

1. Open https://docuprune.com/

2. Choose "Try a sample" and pick a text document, bilingual table or scanned page.

3. Convert it locally, review the Markdown, then try copying or downloading the result. The demo icon in the header shows the workflow.

Which step feels hardest: preserving tables, reading scans, cleaning the result, or exporting a batch? Tell me the format and the step you tried, and what you expected to happen. Please use samples or non-confidential files; do not attach private documents here.

The converter is free and needs no account. Files are processed in your browser; program and reading-model resources may download on first use. OCR can make mistakes, so important text still needs checking against the original.

I am the maker. Specific feedback will help me prioritize improvements.

Launch: https://www.producthunt.com/products/docuprune?launch=docuprune