PDF Narrator

Overview

Transform PDF documents into audiobooks effortlessly using advanced text extraction and text-to-speech technology. This tool is optimized for efficiency, making it ideal for low-VRAM systems and immersive reading experiences.

Demo

Below is a preview of the PDF to Audiobook conversion process:

Features

• 🔍 Intelligent PDF Text Extraction: Removes headers, footers, and page numbers.
• 📖 Chapter or Full-Book Conversion: Extract based on Table of Contents (TOC) or the entire document.
• 🎙️ Customizable Text-to-Speech Settings: Supports multiple TTS models and speaker configurations.
• 💻 Low-Resource Processing: Adjusts chunk sizes dynamically for lower-VRAM systems.
• 🎨 User-Friendly GUI with Theme Customization: Switch themes easily and save preferences.

---

Prerequisites

• Python 3.8+
• FFmpeg: Required for audio processing.
• Piper TTS: For high-quality text-to-speech conversion.

---

Installation

1. Clone the repository

   git clone https://github.com/mateogon/pdf-narrator.git
   cd pdf-narrator

2. Create a virtual environment

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

   pip install -r requirements.txt

4. Install FFmpeg

   • Ubuntu/Debian:

     sudo apt-get install ffmpeg

   • macOS:

     brew install ffmpeg

   • Windows: Download and install from the FFmpeg official site.

5. Download Recommended Voice Model

   • Visit the Hugging Face Piper Voices page.
   • Download the en_US-libritts-high.onnx model and its corresponding en_US-libritts-high.onnx.json file.
   • Place the downloaded files in the models/en/ directory:

     mkdir -p models/en
     mv /path/to/en_US-libritts-high.onnx models/en/
     mv /path/to/en_US-libritts-high.onnx.json models/en/

6. Speaker ID and Voice Samples

   • Use Speaker ID 8 for the recommended voice.
   • Listen to available voice samples for customization options.

---

Quick Start

1. Launch the App

   python main.py

2. Select PDF

   • Browse and select a PDF file.
   • Choose extraction mode: chapters or full book.

3. Configure Audio Settings

   • Select a Piper TTS model.
   • Choose speaker IDs (optional).
   • Adjust chunk size for compatibility with your system.
   • Select audio output format (.wav or .mp3).

4. Generate Audiobook

   • Click Start Process.
   • Monitor progress in real-time with progress bars and logs.

---

Technical Highlights

PDF Extraction

• Automatic Cleaning: Removes headers, footers, page numbers, and excessive whitespace.
• TOC Segmentation: If the document has a TOC, chapters are split accordingly.
• Full-Book Mode: Extracts the entire document if no TOC is available or desired.

Audio Generation

• Chunk-Based Processing: Splits text into manageable chunks for processing.
• Multi-Speaker Support: Customize speaker IDs for varied narration.
• VRAM-Friendly: Dynamically adjusts chunk size based on system performance.

---

Performance Optimization

• Use GPU (CUDA): Select GPU as the processing device for faster TTS generation.
• Adjust Chunk Size: Reduce the chunk size if your system has limited VRAM.
• Experiment with Models: Piper's performance depends on the selected model and speaker IDs.

---

Limitations

• PDF Quality: Extraction accuracy depends on the structure of the source PDF.
• TTS Quality: The quality of the generated audiobook may vary depending on the Piper model.
• Processing Time: Large documents may require significant time to process.

---

Use Cases

• Academia: Convert research papers and textbooks into audiobooks.
• Books: Create audiobooks from eBooks or PDFs.
• Documentation: Narrate user manuals or technical documents.
• Education: Enhance learning with immersive reading.

---

Recommended Workflow

1. Select a PDF file.
2. Choose TOC-based or full-book extraction.
3. Configure Piper TTS settings (model, speaker, chunk size).
4. Generate the audiobook.
5. Use the audiobook for immersive reading or standalone listening.

---

Contributing

Contributions are welcome! If you have ideas for improvements or new features, please open an issue or create a pull request.

---

Acknowledgments

• Piper TTS: High-quality text-to-speech engine.
• PyMuPDF: PDF parsing and text extraction library.
• ttkbootstrap: Modern GUI framework for Python applications.