oomol-lab/pdf-craft is a Python library that converts scanned book PDFs into editable Markdown or easy-to-read EPUB files. In a scanned PDF, the characters remain as images, so the text cannot be handled as text. The repository description says it converts PDFs into several formats, and it lists scanned books, research papers, and technical documents as its targets. The latest release is v2.5.0, and the provided information does not state what changed in this version.
The README explains that the library recognizes body text, chapters, tables of contents, footnotes, tables, formulas, and images. In other words, it does not stop at moving characters over: it also extracts structure such as chapters and footnotes, making the output usable for ebooks and markup. It offers PDF Craft Online, which converts files without any installation, and it supports translating PDFs and EPUBs, generating translated EPUBs and bilingual EPUBs, and writing translations onto the pages of the original PDF. The implementation language is Python, the license is MIT, and the repository has 6,411 stars. The repository was created on February 12, 2025.
That said, no figures are given for the accuracy of structure recognition or for processing speed. From the feature list and the online service alone, it is hard to judge how well a single scanned book will be read. In the score record, only the momentum entry is filled in, at 3.9 on a 10-point scale, while the remaining entries are blank. So the tool looks like an option for people who need to rewrite or search scanned books, but there is not enough evidence to use it right away for work where accuracy matters. The realistic approach is to convert a few pages, check the quality, and then use it.