Cosmisum - Comic/Manga/Manhwa/Manhua Summarizer and Analyzer

A Python pipeline for analyzing manga, comics, and documents. Extracts panels from PDFs, performs OCR, and uses LLM to generate summaries, tags, and genre classification.

Features

• 📄 PDF page extraction and panel detection
• 🔍 OCR text extraction from panels
• 🧠 Smart token-limited chunking with uniform distribution
• 🤖 LLM-powered analysis (OpenAI API compatible)
• 💻 Cross-platform support (Windows, macOS, Linux)

Prerequisites

1. Python 3.10+

Make sure you have Python 3.10 or higher installed.

2. Poppler

Poppler is required for PDF processing. Install it based on your OS:

Windows

Option A: Using Chocolatey (Recommended)

choco install poppler

Option B: Manual Installation

1. Download poppler for Windows from: https://github.com/oschwartz10612/poppler-windows/releases
2. Extract the zip file (e.g., to C:\poppler)
3. Add the bin folder to your PATH:
   • Right-click "This PC" → Properties → Advanced system settings
   • Click "Environment Variables"
   • Under "System variables", find and edit "Path"
   • Add new entry: C:\poppler\Library\bin (or wherever you extracted it)
   • Click OK and restart your terminal

macOS

brew install poppler

Linux (Ubuntu/Debian)

sudo apt-get update
sudo apt-get install poppler-utils

Linux (Fedora/RHEL)

sudo dnf install poppler-utils

3. Tesseract OCR

Tesseract is required for text extraction.

Windows

Option A: Using Chocolatey

choco install tesseract

Option B: Manual Installation

1. Download installer from: https://github.com/UB-Mannheim/tesseract/wiki
2. Run the installer
3. Add Tesseract to PATH (usually C:\Program Files\Tesseract-OCR)
4. Restart your terminal

macOS

brew install tesseract

Linux (Ubuntu/Debian)

sudo apt-get install tesseract-ocr

Linux (Fedora/RHEL)

sudo dnf install tesseract

Installation

1. Clone or download this repository

git clone https://github.com/sammwyy/cosmisum
cd cosmisum

2. Create a virtual environment (recommended but optional)

# Windows
python -m venv venv
venv\Scripts\activate

# macOS/Linux
python3 -m venv venv
source venv/bin/activate

3. Install Python dependencies

pip install -r requirements.txt

Configuration

1. Set up environment variables

On first run, the script will create a default .env file. Edit it with your configuration:

# .env
OPENAI_API_KEY=sk-your-actual-api-key-here
OPENAI_API_BASE_URL=https://api.openai.com/v1
OPENAI_MODEL_ID=gpt-4

Note: You can use any OpenAI-compatible API by changing the OPENAI_API_BASE_URL (e.g., for local models, Azure OpenAI, etc.)

2. Get an API key

• For OpenAI: https://platform.openai.com/api-keys
• For other providers: Check their documentation

Usage

Basic usage

python cosmisum.py input.pdf

Example

python cosmisum.py my_manga_chapter.pdf

What it does

1. Extracts panels from each PDF page
2. Performs OCR on each panel to extract text
3. Creates uniform chunks distributed across the document (respecting token limits)
4. Sends to LLM for analysis
5. Outputs:
   • Plot summary
   • Thematic tags
   • Genre/category classification

Output

The script will display results in the console:

================================================================================
Based on the provided text extracts, here is the analysis of the manga/comic:

1. **Plot Summary**: [Generated summary...]

2. **Thematic Tags**: action, adventure, friendship, ...

3. **Genre/Category**: Shonen manga / Action-Adventure

================================================================================

Troubleshooting

"Unable to get page count. Is poppler installed and in PATH?"

• Make sure poppler is installed (see Prerequisites)
• Verify it's in your PATH: run pdfinfo -v in terminal
• On Windows, restart your terminal after adding to PATH

"pytesseract.pytesseract.TesseractNotFoundError"

• Make sure Tesseract is installed (see Prerequisites)
• On Windows, you may need to specify the path in the script:

  pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'

"OPENAI_API_KEY not configured"

• Edit the .env file and add your actual API key
• Make sure the .env file is in the same directory as cosmisum.py

No text extracted

• Check if your PDF contains actual text (not just images)
• Try adjusting the OCR language settings in the code
• Increase PDF DPI in extract_all_panels() function

Advanced Usage

Using with local LLM

Edit .env to point to your local API:

OPENAI_API_KEY=not-needed
OPENAI_API_BASE_URL=http://localhost:1234/v1
OPENAI_MODEL_ID=local-model

Multi-language OCR

Modify the perform_ocr() function to use different languages:

text = pytesseract.image_to_string(img, lang='jpn+eng')  # Japanese + English

Download language data from: https://github.com/tesseract-ocr/tessdata

License

MIT

Contributing

Pull requests are welcome. For major changes, please open an issue first.