pdf2podcast 0.1.15

A Python library to convert PDF documents into podcasts

🎙️ PDF2Podcast

An advanced Python system for converting PDF documents into audio podcasts with natural dialogues between hosts and experts.

✨ Key Features

• 📄 PDF Extraction: Advanced PDF document processing
• 🤖 LLM Integration: Support for Google Gemini and other models
• 🗣️ Multi-TTS: Amazon Polly, Google TTS, Azure TTS, Kokoro TTS
• 💬 Natural Dialogues: Realistic conversations between host and expert
• 🔍 Semantic Retrieval: Intelligent content selection
• 🎯 Structured Chapters: Automatic organization into thematic chapters
• 🌍 Multilingual: Support for multiple languages
• 🛠️ Modular: Extensible and customizable architecture

🚀 Quick Install

pip install pdf2podcast

📖 Basic Usage

Simple Example

from pdf2podcast import PodcastGenerator
from pdf2podcast.core.rag import AdvancedPDFProcessor

# Setup
processor = AdvancedPDFProcessor()
generator = PodcastGenerator(
    rag_system=processor,
    llm_provider="gemini",
    tts_provider="google",
    llm_config={"api_key": "your-gemini-key"},
    tts_config={"language": "en"}
)

# Generate podcast
result = generator.generate(
    pdf_path="document.pdf",
    output_path="podcast.mp3",
    dialogue=True,  # Dialogue between host and expert
    query="Explain the main concepts of the document"
)

print(f"Podcast generated: {result['audio']['path']}")
print(f"Duration: {result['total_duration']:.1f} seconds")

With Direct Text

# From text instead of PDF
result = generator.generate(
    text="Your text content here...",
    output_path="podcast.mp3",
    dialogue=True,
    query="Discuss the key points"
)

🔧 Advanced Configuration

Semantic Retrieval

from pdf2podcast.core.processing import SimpleChunker, SemanticRetriever

chunker = SimpleChunker()
retriever = SemanticRetriever()

generator = PodcastGenerator(
    rag_system=processor,
    llm_provider="gemini",
    tts_provider="kokoro",
    chunker=chunker,
    retriever=retriever,
    k=5  # Top 5 most relevant chunks
)

Custom Prompts

from pdf2podcast.core.prompts import PodcastPromptBuilder

# Custom instructions
instructions = """
Focus on practical aspects and real-world applications.
Use concrete examples and accessible language.
"""

prompt_builder = PodcastPromptBuilder(
    instructions=instructions,
    dialogue=True
)

generator = PodcastGenerator(
    rag_system=processor,
    llm_provider="gemini",
    tts_provider="azure",
    llm_config={"prompt_builder": prompt_builder}
)

Multi-Voice TTS (Kokoro)

# Different voices for host and expert
generator = PodcastGenerator(
    rag_system=processor,
    llm_provider="gemini",
    tts_provider="kokoro",
    tts_config={
        "voice_id": ["af_heart", "am_liam"],  # [host, expert]
        "language": "en"
    }
)

🎭 Output Modes

Dialogue (Recommended)

• Activation: dialogue=True
• Format: Natural conversation between S1 (host) and S2 (expert)
• Features: Interruptions, questions, clarifications
• Ideal for: Complex content, educational material

Monologue

• Activation: dialogue=False
• Format: Continuous narration
• Features: Linear flow, storytelling
• Ideal for: Narrative content, summaries

🔌 Supported Providers

LLM (Large Language Models)

• Google Gemini ✅ (Recommended)
• OpenAI 🔄 (In development)

TTS (Text-to-Speech)

• Google TTS ✅ - Basic quality, easy setup
• Amazon Polly ✅ - Professional quality
• Azure TTS ✅ - Advanced neural voices
• Kokoro TTS ✅ - Local, multiple voices, precise timing

🛠️ Architecture

📦 pdf2podcast
├── 🎯 PodcastGenerator (Main orchestrator)
├── 📄 RAG Systems (Text extraction)
├── 🧠 LLM Integration (Script generation)
├── 🎵 TTS Engines (Voice synthesis)
├── 🔍 Semantic Retrieval (Intelligent search)
└── 📝 Parser System (Output validation)

Processing Flow

1. Input → PDF/Text
2. RAG → Content extraction and cleaning
3. Chunking → Division into segments
4. Retrieval → Relevant content selection
5. LLM → Structured script generation
6. Parsing → Format validation
7. TTS → Audio conversion
8. Output → Audio file + metadata

📚 Complete Documentation

• 📖 Module Documentation - Detailed architecture
• 💡 Advanced Examples - Use cases and configurations
• 🔧 API Reference - Complete API documentation

⚙️ Environment Setup

Environment Variables

# .env file
GENAI_API_KEY=your_gemini_api_key
AWS_ACCESS_KEY_ID=your_aws_key
AWS_SECRET_ACCESS_KEY=your_aws_secret
AZURE_SUBSCRIPTION_KEY=your_azure_key
AZURE_REGION_NAME=your_azure_region

System Dependencies

# For Kokoro TTS (optional)
pip install torch torchaudio

# For advanced PDF processing
pip install pypdf2 pdfplumber

🎯 Practical Examples

Scientific Paper Podcast

result = generator.generate(
    pdf_path="research_paper.pdf",
    output_path="research_podcast.mp3",
    dialogue=True,
    query="Explain methodology, results and implications",
    instructions="Focus on practical applications and limitations"
)

Technical Documentation Podcast

result = generator.generate(
    pdf_path="technical_manual.pdf",
    output_path="tutorial_podcast.mp3",
    dialogue=True,
    query="Create a step-by-step tutorial",
    instructions="Use concrete examples and common troubleshooting"
)

Multilingual Podcast

# Italian
generator = PodcastGenerator(
    rag_system=processor,
    llm_provider="gemini",
    tts_provider="google",
    llm_config={"language": "it"},
    tts_config={"language": "it"}
)

result = generator.generate(
    text="Italian content here...",
    query="Discuss the main points in Italian"
)

🔍 Troubleshooting

Common Issues

1. API Key Error

# Verify configuration
import os
print("Gemini Key:", os.getenv("GENAI_API_KEY")[:10] + "..." if os.getenv("GENAI_API_KEY") else "Not found")

2. Low Audio Quality

# Use premium TTS provider
tts_config = {
    "voice_id": "Joanna",  # Polly
    "engine": "neural"     # Higher quality
}

3. Short Scripts

# Increase retrieval chunks
generator = PodcastGenerator(..., k=10)  # More content

🤝 Contributing

1. Fork the repository
2. Create feature branch: git checkout -b feature/new-feature
3. Commit: git commit -am 'Add new feature'
4. Push: git push origin feature/new-feature
5. Create Pull Request

📜 License

MIT License - see LICENSE for details.

🙏 Credits

• LangChain - LLM Framework
• Pydantic - Data validation
• Sentence Transformers - Semantic embeddings
• Community Contributors - Feedback and improvements

---

🚀 Roadmap

• OpenAI Integration - GPT-4 support
• Batch Processing - Multiple file processing
• Web Interface - Web-based GUI
• Audio Effects - Background music, effects
• Export Formats - MP3, WAV, OGG
• Cloud Deployment - Docker, AWS Lambda

Last updated: December 2024 Version: 1.0.0