Book Extract

A collection of UGLY, but functional applications to automate the process of digitizing physical books into EPUB and M4B audiobook formats. This toolchain provides interfaces for capturing, processing, OCR, editing, and rendering steps to convert books from screenshots to structured e-books and audiobooks.

Overview

This project provides a complete pipeline with intuitive GUI applications for:

1. Capturing page screenshots from a displayed book with automated navigation
2. OCR Processing to extract and clean text using Tesseract OCR and AI-powered correction
3. Editing the processed content and generating formatted JSON files
4. Rendering the final EPUB or M4B audiobook with proper formatting and metadata

Prerequisites

• Linux system with X11 (for screenshot and mouse automation)
• Python 3 interpreter
• Tesseract OCR
• ImageMagick (import command)
• xdotool (for mouse automation)
• API access to a compatible AI model (supports OpenAI, Anthropic, OpenRouter, etc.)
• For M4B audiobook generation:
  • Kokoro TTS (text-to-speech engine)
  • FFmpeg (audio processing)
  • FFprobe (media analysis)

Installation

1. Clone this repository
2. Install required system dependencies:

   sudo apt-get install python3 python3-tk tesseract-ocr imagemagick xdotool ffmpeg python3-venv

3. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

4. Install Python dependencies:

   pip install -r requirements.txt

5. Create a .env file based on the example:

   cp .env-example .env

   Then edit .env with your API credentials:

   API_TOKEN="your_api_key_here"
   MODEL="anthropic/claude-3.7-sonnet"
   API_URL="https://openrouter.ai/api/v1/chat/completions"
   

Note: The kokoro package provides the TTS engine, while ffmpeg includes both FFmpeg and FFprobe for audio processing and media analysis. ImageMagick provides the import command used for screenshot capture.

Quick Start

1. Setup: Install dependencies and configure your .env file with API credentials
2. Capture: Run python3 capture_gui.py to capture and crop pages
3. OCR: Run python3 ocr_gui.py to extract and clean text with Tesseract OCR and AI
4. Edit: Run python3 edit_gui.py to edit content, preview with image rendering, and generate formatted JSON files
5. Export: Run python3 render_epub.py for EPUB or python3 render_m4b.py for M4B audiobook

Usage

The digitization process follows these steps in sequence using the GUI applications:

1. Capture & Crop

Use the caputure and crop tool to record cropped screenshots:

python3 capture_gui.py

The Capture GUI provides:

• Image Preview: Take test screenshots and preview crop areas with interactive selection
• Real-time Crop Selection: Click and drag to select crop coordinates with visual feedback
• Sequence Numbering: Set initial sequence numbers for captured images
• Progress Tracking: Real-time status updates for the capture and crop process
• Coordinate Testing: Test mouse coordinates before starting capture
• All-in-One Interface: Single tool replacing separate capture and crop steps

2. Extract Text with OCR

Extract text from cropped images using OCR and AI processing:

python3 ocr_gui.py

The OCR GUI provides:

• Complete OCR pipeline with three processing stages:
  1. Tesseract OCR: Initial text extraction from images
  2. AI Second Pass: LLM-powered OCR correction and improvement
  3. AI Final Pass: Content cleanup and merging across page breaks
• Configurable API settings with connection testing
• Real-time progress tracking for batch processing
• Results preview functionality including final merged content
• Support for basic OCR-only or full pipeline processing
• AI processing to correct OCR mistakes and structure content

3. Edit and Format Content

Edit the processed content and generate formatted JSON files:

python3 edit_gui.py

The edit GUI provides:

• Rich text editor for reviewing and editing extracted content
• JSON structure editing with validation
• Real-time preview of formatted content with actual image rendering
• Metadata editing (title, author, chapters, etc.)
• Export to intermediate JSON format for rendering
• Image preview and management with support for PNG, JPEG, GIF, and BMP formats
• Content organization and chapter structuring
• Cover image and content image integration

4. Export Final Format

Generate the final e-book or audiobook from the formatted JSON:

For EPUB E-books:

python3 render_epub.py

For M4B Audiobooks:

python3 render_m4b.py

Both export tools provide:

• Loading of intermediate JSON format files
• Metadata configuration and validation
• Progress tracking during generation
• Preview capabilities
• Professional formatting with proper chapters and structure

Intermediate JSON Format

BookExtract uses a unified intermediate JSON representation that bridges the gap between OCR processing and final output generation. This format provides:

• Unified Structure: Single format that works with both EPUB and M4B generation
• Rich Metadata: Comprehensive book information including title, author, language, etc.
• Structured Content: Hierarchical organization with chapters and typed content sections
• Image Support: Native support for cover images and content images with captions
• Format Conversion: Seamless conversion between different processing stages
• Enhanced Features: Word counting, content analysis, and extensible design

Image Integration in JSON Format

The intermediate JSON format supports two types of images:

Cover Images

{
  "type": "cover",
  "image": "cover.png"
}

Content Images

{
  "type": "image",
  "image": "photo.jpg",
  "caption": "Optional caption text"
}

Supported Image Formats: PNG, JPEG, GIF, and BMP File Placement: Images should be placed in the same directory as the JSON file or use relative paths Resolution Guidelines:

• Cover images: Recommended 600x800 pixels or similar book cover ratio
• Content images: Any reasonable resolution (automatically resized for display)

Customization

• Capture Settings: Use the capture GUI to configure page count, navigation coordinates, and timing
• Crop Dimensions: Use the crop GUI's interactive preview to set optimal crop areas for your display
• OCR Processing: Configure API settings, models, and processing options through the OCR GUI
• Content Editing: Use the edit GUI to refine extracted text, organize chapters, and adjust metadata
• Export Options: Configure output settings in the render GUIs for EPUB and M4B formats
• AI Prompts: Modify the prompts within ocr_gui.py to improve content structuring for specific book types

Troubleshooting

Image Rendering Issues

• Image Not Displaying: Check that the image file exists in the specified path and verify the format is supported (PNG, JPEG, GIF, BMP)
• Performance Issues: Use the "Refresh" button to clear the image cache, or consider resizing very large images
• Error Messages:
  • [IMAGE: filename - NOT FOUND]: File doesn't exist at the specified path
  • [IMAGE: filename - LOAD FAILED]: File exists but couldn't be loaded (possibly corrupted or unsupported format)

General Issues

• Ensure all dependencies are properly installed, especially Pillow (PIL) for image processing
• Check file permissions for image files and directories
• For memory issues with large images, close and reopen the tool

Limitations

• This toolchain is experimental and may require adjustments for specific books or layouts
• OCR quality depends on the clarity of source material and image resolution
• Complex formatting elements (tables, multi-column layouts) may not be perfectly preserved
• Currently optimized for left-to-right reading and standard book layouts
• Requires manual setup of capture coordinates for each book viewer application
• Image files should be kept under 5MB for optimal performance

License

This project is licensed under the GNU General Public License v3.0 - see the LICENSE file for details.