Let me paint you a picture, You have a scanned book. Maybe it’s an old technical manual. Maybe it’s a collection of research papers. Maybe it’s that textbook from 1998 that never existed in digital form.
You need to work with the content. You need to search it, quote it, maybe even convert it to EPUB so you can read it on your e-reader.
But here’s the problem: it’s a scanned PDF. The text isn’t selectable. The tables are images. The formulas look like hieroglyphs. And every PDF tool you’ve tried either gives you garbage output or takes forever.
That’s exactly why PDF-Craft is created.
What PDF-Craft Actually Does
PDF-Craft is an open-source tool that converts PDF files, especially tricky scanned books, into clean Markdown or EPUB formats.
The secret sauce? It uses DeepSeek OCR for document recognition. That means it can actually understand what’s on the page. Tables? Handled. Formulas? Preserved. Headers and footers? Filtered out automatically.
You feed it a PDF. It gives you back usable, readable content.
The Best Part? It Runs Locally
Here’s what I love about this tool.
You can run everything on your own machine. No uploading sensitive documents to some cloud server. No waiting in a queue. No network failures halfway through a 400-page book.
With GPU acceleration, PDF-Craft processes documents entirely offline. The old version used to rely on LLMs for text correction, which meant network calls and long waits. The new v1.0.0 embraces DeepSeek OCR directly and finishes the job locally.
Unlike other OCR apps, this app is faster, and private.
What You Can Do With PDF-Craft
1- You can convert scanned books to Markdown.
Perfect for researchers who want to quote from physical books. The app is ideal for writers who need to extract passages.
This is perfect for anyone who wants their PDF content in a format they can actually edit.
2- You can generate EPUB files with automatic tables of contents.
If you’ve ever tried to read a scanned PDF on an e-reader, you know the pain. PDF-Craft handles this gracefully. The app identifies document structure, extracts the body text, and builds a proper EPUB with working navigation.
3- You can preserve complex content.
Academic papers with footnotes? Handled. Technical documents with formulas? Preserved. Tables that would confuse most converters? PDF-Craft keeps them intact. Even images inside footnotes are maintained.
4- You can process documents offline.
No internet required. No privacy concerns. Your documents stay on your machine.
Try It Before You Install
You can try Inkora – PDF Craft, an online version built around the same workflow. Upload a PDF, see how it works, and decide if it fits your needs.
It’s the same engine, just running in your browser.
Install
You need Python and PIP to install:
pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
pip install pdf-craftA Quick Note on Versions
The current v1.0.0 is fast, local, and fully powered by DeepSeek OCR. It removed the LLM text correction feature for better performance.
If you still need that text correction functionality, you can stick with v0.2.8. Both versions are available, so you can choose what works for your workflow.
Who This Is For
- Researchers who need to extract clean text from scanned academic papers.
- Writers who want to reference physical books in their digital workflow.
- Students who need to convert textbooks to EPUB for reading on the go.
- Archivists who are digitizing old documents and need clean output.
- Anyone who has ever stared at a scanned PDF and wished they could just copy the text.
Getting Started
PDF-Craft is open source. You can find it on GitHub, clone it, and start converting.
Or visit Inkora to try the online version first.
Either way, you get the same core experience: fast, local, privacy-friendly PDF conversion that actually understands what’s on the page.
No more wrestling with scanned PDFs. No more garbage output. No more waiting for cloud services to fail.
Just clean Markdown and EPUB files, ready to use.
Have you tried PDF-Craft?
What kind of documents are you converting?
Drop your experience in the comments.
License
MIT License
