VoiceSlide

🎙️ NLP-powered presentations with voice navigation, live highlighting, and speech analytics.

---

📖 Table of Contents

• Overview
• Key Features
• Prerequisites
• Installation
• Usage
• Configuration
• Project Structure
• Contributing
• License
• Contact & Acknowledgements

---

🔭 Overview

VoiceSlide is a full-stack presentation platform that lets you control slides with your voice. It uses on-device speech-to-text (faster-whisper), semantic intent classification, and voice activity detection (Silero VAD) to deliver a hands-free presenting experience — no cloud APIs, no microphone button mashing, no latency.

Why? Traditional presentation tools force speakers to click, tap, or use a clicker. VoiceSlide replaces all of that: just speak naturally, and the system navigates to the right slide, highlights keywords in real time, answers your Q&A from speaker notes, and tracks your speaking analytics — all locally and in real time.

How? A Flask + Socket.IO backend streams browser microphone audio through a VAD → Whisper → NLP pipeline. The frontend renders slides with reveal.js and reacts to WebSocket events for navigation, highlighting, and analytics.

---

✨ Key Features

Feature	Description
🗣️ Voice Navigation	Say "next slide", "go to slide 5", or even describe content — the system finds and navigates to the right slide using semantic search.
🎯 Live Keyword Highlighting	As you speak, matching words on the current slide are highlighted in real time using fuzzy matching.
🧠 Semantic Content Search	Uses sentence-transformer embeddings to match spoken phrases to slide content, even when wording differs.
🎤 Voice Activity Detection	Silero VAD detects speech boundaries — transcription only fires when you actually pause, eliminating false triggers.
📊 Speech Analytics Dashboard	Post-presentation dashboard with filler word tracking, words-per-minute, and VADER sentiment analysis over time.
💬 Q&A from Speaker Notes	Ask a question during your talk and the system searches your speaker notes for relevant answers, displayed on the Presenter Panel.
📑 PPTX Import	Upload a .pptx file and it's automatically converted to VoiceSlide's slide format — no manual JSON editing required.
🖥️ Presenter Panel	A private second-screen view with speaker notes, Q&A results, and current slide context.
🌐 Fully Local	All NLP runs on-device. No cloud APIs, no data leaves your machine.

---

📋 Prerequisites

Requirement	Version	Notes
Python	3.10+	Required for type hints and library compatibility
PyTorch	2.x	With CUDA 12.1 for GPU acceleration (CPU works but is slower)
torchaudio	2.x	Required by Silero VAD
Node.js	—	Not required — frontend uses CDN-loaded libraries
FFmpeg	—	Not required — faster-whisper handles raw PCM directly

Note

PyTorch and torchaudio must be installed manually for your CUDA version before running pip install. See Installation below.

---

🚀 Installation

1. Clone the Repository

git clone [INSERT_REPO_URL].git
cd voiceslide

2. Create a Virtual Environment

python -m venv venv
source venv/bin/activate        # Linux / macOS
venv\Scripts\activate           # Windows

3. Install PyTorch (GPU)

Install PyTorch and torchaudio for your CUDA version. Example for CUDA 12.1:

pip install torch torchaudio --index-url https://download.pytorch.org/whl/cu121

Tip

For CPU-only: pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu

4. Install Python Dependencies

pip install -r backend/requirements.txt

5. Add Slide Content

Either upload a .pptx through the web UI or place a slides.json file in the data/ directory:

mkdir -p data
# Option A: Start the server and use /upload in the browser
# Option B: Create data/slides.json manually (see Usage below)

6. Start the Server

python backend/app.py

The server starts on http://localhost:5000 by default.

---

🎮 Usage

Presenting with Voice Control

1. Open http://localhost:5000 in your browser.
2. Click the microphone button (bottom-right) to enable voice input.
3. Speak naturally — VoiceSlide handles the rest:

"Next slide"           → advances one slide
"Previous slide"       → goes back one slide
"Go to slide 3"       → jumps to slide 3
"Show the revenue chart" → semantic search finds the matching slide
"First slide"          → jumps to the beginning
"Last slide"           → jumps to the end

Presenter Panel

Open http://localhost:5000/presenter in a second browser window (or second monitor) to see:

• Current slide speaker notes
• Real-time Q&A results from your notes
• Transcript feed

Speech Analytics

After your presentation, open http://localhost:5000/analytics to review:

• Average WPM — were you rushing or dragging?
• Filler word breakdown — how many "um"s, "like"s, "you know"s?
• Sentiment timeline — was your language confident and positive?

Slide Data Format

VoiceSlide uses a simple JSON format. You can create slides manually or import a .pptx:

{
  "slides": [
    {
      "title": "Welcome",
      "content": "<h1>Welcome to VoiceSlide</h1><p>Hands-free presentations.</p>",
      "notes": "Introduce the product and greet the audience."
    },
    {
      "title": "Revenue Growth",
      "content": "<h2>Revenue</h2><p>Revenue grew 18% year over year.</p>",
      "notes": "Key talking point: 18% growth driven by enterprise segment."
    }
  ]
}

---

⚙️ Configuration

All configuration is managed through environment variables. Defaults are defined in backend/config.py.

Variable	Default	Description
VOICESLIDE_HOST	0.0.0.0	Server bind address
VOICESLIDE_PORT	5000	Server port
VOICESLIDE_DEBUG	true	Enable Flask debug mode

# Example: Run on port 8080 with debug off
export VOICESLIDE_HOST="127.0.0.1"
export VOICESLIDE_PORT="8080"
export VOICESLIDE_DEBUG="false"
python backend/app.py

---

📁 Project Structure

voiceslide/
├── backend/
│   ├── app.py                 # Flask + Socket.IO entry point
│   ├── config.py              # Environment-based configuration
│   ├── transcriber.py         # faster-whisper STT engine
│   ├── vad_engine.py          # Silero VAD speech detection
│   ├── intent_classifier.py   # Voice command classification
│   ├── context_search.py      # Semantic slide search (sentence-transformers)
│   ├── keyword_highlighter.py # Fuzzy keyword matching for live highlights
│   ├── qa_assistant.py        # Q&A from speaker notes
│   ├── analytics.py           # Speech analytics (fillers, WPM, sentiment)
│   ├── slide_loader.py        # JSON slide loading & validation
│   ├── pptx_converter.py      # .pptx → slides.json converter
│   └── requirements.txt       # Python dependencies
├── frontend/
│   ├── index.html             # Main presentation view
│   ├── upload.html            # Slide upload / editor page
│   ├── presenter.html         # Presenter Panel (speaker notes + Q&A)
│   ├── analytics.html         # Speech Analytics Dashboard
│   ├── css/
│   │   ├── style.css          # Global styles & design tokens
│   │   ├── presentation.css   # Slide presentation styles
│   │   ├── presenter.css      # Presenter Panel styles
│   │   ├── upload.css         # Upload page styles
│   │   └── analytics.css      # Analytics Dashboard styles
│   └── js/
│       ├── app.js             # Main presentation logic + WebSocket
│       ├── presenter.js       # Presenter Panel logic
│       ├── upload.js          # Upload page logic
│       ├── analytics.js       # Analytics Dashboard charts (Chart.js)
│       └── audio-processor.js # AudioWorklet for mic capture
├── tests/
│   ├── test_analytics.py
│   ├── test_context_search.py
│   ├── test_intent_classifier.py
│   ├── test_interceptor.py
│   ├── test_keyword_highlighter.py
│   ├── test_qa_assistant.py
│   ├── test_transcriber.py
│   ├── test_universal_fallback.py
│   └── fixtures/
│       ├── sample.pptx
│       └── sample_audio.raw
└── data/
    └── slides.json            # Active slide content (auto-generated)

---

🤝 Contributing

Contributions are welcome! Here's how to get started:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Development Guidelines

• Follow the existing code style — no comments unless they match existing patterns or explain complex logic.
• All new backend features must include tests in tests/.
• Run the full test suite before submitting:

pytest tests/ -v

Important

PyTorch and torchaudio are installed manually and are not listed in requirements.txt. Make sure your environment has them installed before running tests.

---

📄 License

Distributed under the [INSERT_LICENSE] License. See LICENSE for more information.

---

💬 Contact & Acknowledgements

[Muhammad Hassan] — [muhammadhassan1762005@gmail.com] — @MY_LinkedIn

Project Link: [INSERT_REPO_URL]

Built With

• Flask — lightweight Python web framework
• Socket.IO — real-time bidirectional communication
• faster-whisper — CTranslate2-based Whisper inference
• Silero VAD — voice activity detection
• sentence-transformers — semantic text embeddings
• VADER Sentiment — lexicon-based sentiment analysis
• reveal.js — HTML presentation framework
• Chart.js — JavaScript charting library
• thefuzz — fuzzy string matching

---

Made with ❤️ and a whole lot of voice commands.