README.md

metadata

title: PDF Extractor
emoji: 📄
colorFrom: blue
colorTo: indigo
sdk: docker
app_port: 7860
pinned: false
license: mit
short_description: Extract text from typed, scanned, and handwritten PDFs.

PDF Extractor

A live demo of a local-first PDF text extractor. Drop in a PDF and it returns a clean, searchable, exportable transcript. Pages that already have a text layer are read directly; scanned image pages are run through OCR.

This Space runs on a free shared CPU, so OCR is slower than it would be on your own machine and there can be a short cold-start delay when the Space wakes up. The first request after a rebuild may also be slow while model weights load.

Source code and instructions for running it locally: https://github.com/Archie0099/PDF-Extractor

Notes

• Typed or born-digital pages are extracted instantly and exactly, with no OCR.
• Scanned pages use PaddleOCR (English and Hindi).
• For handwriting, the optional online OCR path (Google Gemini) is far more accurate. It is off by default and needs your own free API key, which stays in your browser.
• Nothing is stored. Uploaded files are held in memory only while a job runs and are dropped afterward.