Formats
What it reads, what comes out, and how we know.
One guide per format. Each says what is accepted, what the output looks like, how the code gets there, where each claim comes from, and what it does not do.
The guides.
- HTML Saved or fetched HTML to Markdown, with page furniture removed and tables kept.
- WARC Capture pages to WARC, extract text from archives and replay records as one HTML file.
- PDF Text, headings, tables and images from born-digital PDFs, read in process.
- OCR Text from scanned PDF pages, through Poppler and Tesseract run on your machine.
- DOCX In Webfold's main branch; not in the 0.1.0 release. Word files (DOCX, DOCM) to Markdown: headings from styles, lists, links, tables, footnotes.
- PPTX In Webfold's main branch; not in the 0.1.0 release. PowerPoint (PPTX) to Markdown: slides in presentation order, with titles, text, tables and notes.
- RTF In Webfold's main branch; not in the 0.1.0 release. RTF in and out: a structured reader, Markdown and plain text, and an ASCII writer.
- Spreadsheets XLSX and XLS fixes are in Webfold's main branch; not in the 0.1.0 release. XLSX, XLSB, ODS, CSV and legacy XLS to Markdown tables and structured cell rows.
- OpenDocument ODT, ODP and ODS to Markdown, and the structure the reader drops.
- EPUB EPUB books to Markdown, with two reading orders depending on the route.
- Scientific sources In superdocuments and Webfold's pin; not on the Webfold command line. LaTeX, BibTeX, arXiv source archives and MathML, read as structure.
See it
What comes out of real files?
Each result below is what the webfold command printed for that file, from the current development build. Headings, lists, links, tables, dates, notes and sheet names survive. The clutter does not.
In · page.html
A page with a nav bar, a cookie banner, an article with a list and a table, an ad block and a footer.
Out · Markdown
# Quarterly update
Revenue grew **12%** on the quarter. See the [full report](/reports/q3).
## Highlights
- New region launched
- Churn down to 1.9%
| Region | Revenue |
| --- | --- |
| EMEA | 4.2M |
| APAC | 3.1M |$ webfold scrape page.html
How to read these pages.
Every claim on a guide has a row in its evidence table. Read from the code means the behaviour is in the implementation at the path given. Covered by a test names a test in the repository; the guide does not claim it was run for the current release. Measured points to a benchmark with its protocol and limits.
Where a capability is in the main branch and not in a released build, the guide says so at the top. Where we found a defect while writing a guide, the limit says what it does today, not what we intend.
Not covered yet
- PDF and OCR quality is scored on small samples of public test sets. Treat it as a first pass.
- Competitor comparisons appear only where a dated record exists; most formats have none yet.
- Hosted API behaviour for documents and OCR is not described here, because it cannot be checked from the open code.
Try it on your own files.
Run the command-line tool on your documents and compare against the guide.