Hack the HydraKey facts
| Input formats | PDF with selectable text (scanned PDFs need OCR PDF) |
|---|---|
| Size limit | 100 MB per file (checked in the browser) |
| Where it runs | In your browser with pdf.js 4.10.38; the file is not uploaded (the library itself is downloaded from a CDN) |
| Cost | Free, no sign-up |
| Output | Markdown text (headings, paragraphs, lists), copy or download as .md |
How headings are found
The converter measures the font size of every line. The most common size is treated as body text; lines clearly larger become headings: about 1.6 times the body size gives #, 1.3 times ##, and 1.12 times ###. Bold text at body size stays normal text, so a PDF whose titles are only bold may come out without headings: add # by hand where needed. Lines of the same paragraph are joined and hyphenated words at line ends are repaired.
What to check
Lines are read from the top of each page to the bottom, which suits most reports, contracts and manuals. Pages with two or more columns side by side can come out mixed, so check them. Tables keep their text row by row but do not become Markdown tables; images are not included. If a file gives almost no text, it is a scan: the tool tells you, and OCR PDF can read it.
Example
A one-page PDF with a title, two section headings and body text. Larger, bold lines become headings and the line break inside the paragraph is joined.
Input
(one-page PDF) Remote work policy <- 20 pt bold Who can work remotely <- 14 pt bold Employees can work from home up to three days a week, after agreement with their manager. <- 11 pt, two lines Equipment <- 14 pt bold The company provides a laptop and a headset.
Output
# Remote work policy ### Who can work remotely Employees can work from home up to three days a week, after agreement with their manager. ### Equipment The company provides a laptop and a headset.