ocr-document

description	Extracts text and tables from PDFs and images using PyMuPDF, pdfplumber, and Tesseract OCR (Norwegian/English), with auto/text/ocr/hybrid methods and page selection. Use when reading, OCR-ing, or extracting text/tables from a PDF, scanned document, or image.
name	ocr-document
triggers	["PDF","OCR","document","extract text","scan","read document"]
category	media
maturity	stable
tags	["ocr","tesseract","pdf-extraction","pymupdf","norwegian"]

OCR Document Skill

Extract text from PDF/image

SKILL=${CLAUDE_SKILLS_DIR:-$HOME/.claude-agent/.claude/skills}/ocr-document/scripts

# Auto-detect best method (text vs OCR)
python3 $SKILL/extract.py document.pdf

# Force OCR (for scanned documents)
python3 $SKILL/extract.py document.pdf --method ocr --lang nor+eng

# Extract tables (pdfplumber)
python3 $SKILL/extract.py document.pdf --method pdfplumber

# Specific pages
python3 $SKILL/extract.py document.pdf --pages 1-5

# Save to file
python3 $SKILL/extract.py document.pdf --output extracted.md

# OCR an image
python3 $SKILL/extract.py photo.jpg --lang nor+eng

Download Telegram files

# Download by file_id
python3 $SKILL/telegram_file.py --file-id <id> --output file.pdf

# Get most recent document from a chat
python3 $SKILL/telegram_file.py --chat-id <your-telegram-user-id> --recent --output file.pdf

Methods

• auto: Try text extraction first, fall back to OCR if low text content
• text: PyMuPDF direct extraction (fastest)
• pdfplumber: Better for tables and structured data
• ocr: Tesseract OCR via pdf2image (for scanned docs)
• hybrid: Text first, OCR only for pages that need it

Languages

Default: nor+eng (Norwegian + English). Change with --lang. Available: tesseract --list-langs