Questions

Yes — this is the core of what Paragra does. The output is a real, reflowable EPUB with proper chapters and structure, built to be read like any other book, not a fixed-page PDF wrapped in a new file extension.

Yes. Chapters are detected automatically, and the EPUB's table of contents is generated from them — no manual entry needed. Prefer a different chapter title? The built-in editor lets you rename any chapter before you export, so the table of contents always reads the way you want.

Yes. Lines that are just a page number, and lines that repeat identically across many pages (typical of running headers/footers), are detected and stripped before a chapter's content is built.

Yes. Words broken across a line by a hyphen are rejoined, and Apple Intelligence, running entirely on your Mac, restores natural paragraph breaks from the raw scanned text.

Not beyond reading lines in order. Paragra runs on-device OCR and an AI cleanup pass for paragraphs, but there's no dedicated analysis of a page's regions or reading order today.

No — every page is OCR'd, even ones that already contain a real text layer. Detecting existing text and skipping OCR for it is a possible future improvement.

No. Table structure isn't detected — rows and columns on a scanned page currently come through as plain, unstructured text.

No. There's no LaTeX or MathML support — mathematical notation is currently recognized as plain text, like any other character.

Not reliably. Text is read in the order OCR returns it, without column-aware sorting — a two-column page can come out with its reading order scrambled.

No. Paragra only distinguishes chapter headings from body paragraphs today — it doesn't yet classify or style blockquotes, code blocks, definitions, or similar special content.