PDF to Speech

A full-stack application that converts PDF documents into natural-sounding speech using AI-powered text-to-speech (TTS) technology. Upload a PDF, and listen to it read aloud with professional voice synthesis powered by Supertonic.

🌟 Features

• PDF Upload & Parsing: Extract text from any PDF document
• Intelligent Text Processing: Automatically split text into sentences using NLP
• High-Quality TTS: Generate natural-sounding speech using Supertonic's AI voice models
• Interactive Player: Play, pause, resume, and navigate through sentences
• Visual Feedback: Highlight the currently playing sentence with auto-scroll
• Seamless Experience: Modern, responsive UI built with Next.js and Material-UI

🏗️ Architecture

The application consists of two main components:

Backend (Python/FastAPI)

• PDF Parsing: Uses pdfminer.six to extract text from PDF files
• NLP Processing: Leverages spaCy (en_core_web_sm) to split text into sentences
• Text-to-Speech: Integrates Supertonic's ONNX models for voice synthesis
• API Endpoints:
  • POST /api/upload - Upload PDF and extract sentences
  • POST /api/tts - Synthesize speech for a given sentence

Frontend (Next.js/React)

• Modern UI: Built with Next.js 14, React 18, and Material-UI 6
• Component-Based: Modular architecture with reusable components
• Interactive Controls: Player controls for playback management
• Static Export: Optimized for deployment as static files

📋 Prerequisites

• Docker and Docker Compose (recommended)
• OR:
  • Python 3.11+
  • Node.js 20+
  • Git LFS (for downloading TTS models)

🚀 Quick Start

Using Docker (Recommended)

1. Clone the repository:

   git clone <repository-url>
   cd document-reader-supertonic

2. Build and run:

   docker build -t pdf-tts .
   docker run -p 8000:8000 pdf-tts

3. Access the application: Open your browser to http://localhost:8000

Manual Setup

Backend Setup

1. Install Python dependencies:

   cd backend
   pip install -r requirements.txt

2. Download spaCy model:

   pip install https://github.com/explosion/spacy-models/releases/download/en_core_web_sm-3.7.1/en_core_web_sm-3.7.1.tar.gz

3. Download Supertonic models:

   git lfs install
   git clone https://huggingface.co/Supertone/supertonic /models/supertonic

4. Download helper script:

   curl -o app/helper.py https://raw.githubusercontent.com/supertone-inc/supertonic/main/py/helper.py

5. Run the backend:

   uvicorn app.main:app --host 0.0.0.0 --port 8000

Frontend Setup

1. Install dependencies:

   cd frontend
   npm install

2. Run development server:

   npm run dev

3. Or build for production:

   npm run build

📖 Usage

1. Upload a PDF: Click the "Upload PDF" button and select a PDF file
2. View Sentences: The extracted text appears in the viewer, split into sentences
3. Play Audio: Click "Play" to start listening from the beginning or current sentence
4. Navigate: Use "Next" and "Prev" buttons to skip between sentences
5. Jump to Sentence: Double-click any sentence to start playing from that point
6. Auto-Scroll: Toggle auto-scroll to automatically follow along with the audio
7. Playback Controls: Use Play, Pause, Resume, and Stop buttons as needed

🛠️ Technology Stack

Backend

• FastAPI: Modern, fast web framework for building APIs
• PDFMiner.six: Robust PDF text extraction
• spaCy: Industrial-strength NLP for sentence segmentation
• Supertonic: State-of-the-art neural TTS with ONNX runtime
• ONNX Runtime: Efficient model inference
• Uvicorn: Lightning-fast ASGI server

Frontend

• Next.js 14: React framework with static export capabilities
• React 18: Component-based UI library
• Material-UI (MUI) 6: Comprehensive component library
• TypeScript: Type-safe JavaScript development
• Emotion: CSS-in-JS styling solution

📁 Project Structure

document-reader-supertonic/
├── Dockerfile              # Multi-stage Docker build
├── backend/
│   ├── requirements.txt    # Python dependencies
│   └── app/
│       ├── main.py         # FastAPI application & routes
│       ├── pdf_parser.py   # PDF text extraction
│       ├── nlp.py          # Sentence segmentation
│       └── tts.py          # Text-to-speech synthesis
└── frontend/
    ├── package.json        # Node.js dependencies
    ├── next.config.js      # Next.js configuration
    └── src/
        ├── components/     # React components
        │   ├── PdfUploader.tsx     # File upload component
        │   ├── PlayerControls.tsx  # Audio player controls
        │   └── TextViewer.tsx      # Sentence display & navigation
        ├── lib/
        │   └── api.ts      # API client functions
        └── pages/
            └── index.tsx   # Main application page

🔧 Configuration

Voice Styles

The default voice style is set to M1.json (male voice). You can modify the voice style in backend/app/tts.py:

VOICE_STYLE = ["/models/supertonic/voice_styles/M1.json"]

Available voice styles are located in /models/supertonic/voice_styles/.

TTS Parameters

Adjust synthesis parameters in the SupertonicTTS.synthesize() method:

• total_step: Number of diffusion steps (default: 5)
• speed: Playback speed multiplier (default: 1.05)

🐳 Docker Details

The Dockerfile uses a multi-stage build:

1. Frontend Stage: Builds the Next.js application into static files
2. Backend Stage: Sets up Python environment, installs dependencies, downloads models, and serves both API and static frontend

📝 API Reference

POST /api/upload

Upload a PDF file and extract sentences.

Request: multipart/form-data with file field

Response:

{
  "sentences": [
    { "id": 0, "text": "First sentence." },
    { "id": 1, "text": "Second sentence." }
  ]
}

POST /api/tts

Synthesize speech for a given text.

Request:

{
  "text": "Text to synthesize"
}

Response:

{
  "audioUrl": "/data/audio/tmp_xyz.wav"
}

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

This project uses the following third-party technologies:

• Supertonic - Text-to-speech model
• spaCy - Natural language processing
• FastAPI - Web framework
• Next.js - React framework

🙏 Acknowledgments

• Supertone for providing the high-quality TTS models
• spaCy for robust NLP capabilities
• The open-source community for all the amazing tools and libraries

📧 Contact

For questions, issues, or suggestions, please open an issue on the repository.