ocr-research

name	ocr-research
description	Extract text from any PDF — text-based, scanned, or with embedded images (receipts, photos). Local OCR using PyMuPDF + Tesseract. Supports English, Hindi, Sanskrit.

OCR Research Documents

Extract text from PDFs locally using PyMuPDF (text extraction) and Tesseract OCR (scanned pages + embedded images).

Quick Start

# Install dependencies (auto-installs if missing)
cd ~/.agents/skills/ocr-research
bash bin/setup.sh

# Extract text from any PDF
ocr bin/ocr <pdf_path>

# With options
ocr bin/ocr paper.pdf --pages 1-10
ocr bin/ocr receipt.pdf --lang eng
ocr bin/ocr sanskrit_doc.pdf --lang eng+san
ocr bin/ocr doc.pdf --dpi 400 --output output.txt

What It Handles

PDF Type	Method	Speed
Text-based PDFs	PyMuPDF extraction	Instant
Scanned pages	Render → Tesseract OCR	~2-5s/page
Embedded images	Extract each image → OCR	~1-3s/image
Mixed content	Smart combo of all above	Varies

Receipts, invoices, photos-in-PDFs, scanned academic papers — all work.

Languages

ocr bin/ocr doc.pdf --lang eng          # English
ocr bin/ocr doc.pdf --lang eng+hin      # English + Hindi
ocr bin/ocr doc.pdf --lang eng+san      # English + Sanskrit (Devanagari)

Options

--output, -o FILE    Write to file instead of stdout
--pages, -p RANGE    Page range: "1-5" or "1,3,7"
--lang, -l LANG      Tesseract language(s) (default: eng)
--dpi, -d N          DPI for rendering (default: 300, use 400+ for small text)
--verbose, -v        Show progress

Setup (one-time)

The bin/setup.sh script installs:

• Python 3 (via system or brew)
• Tesseract 5 + language packs (brew)
• PyMuPDF in a local venv

bash bin/setup.sh

Examples

# Receipt from banking app
ocr bin/ocr DG_Oil_Payment.pdf

# Research paper with mixed content
ocr bin/ocr paper.pdf --lang eng+san

# Extract just pages 3-7
ocr bin/ocr report.pdf --pages 3-7

# High quality for small print
ocr bin/ocr flyer.pdf --dpi 400 --output extracted.txt