real-time-speech-analyzer

Real-Time Speech Analyzer

🎙️ Real-time speech analysis with local LLM using multiple concurrent analysis instructions

Analyze speech in real-time using Web Speech API and local Ollama models. Run multiple analysis instructions simultaneously on each speech segment with live processing feedback.

✨ Features

• Real-time speech transcription with live word-by-word display
• Multiple analysis instructions running concurrently on each speech chunk
• Conversation context management with rolling summaries
• Local LLM integration via Ollama (privacy-focused, no cloud required)
• Processing notifications with live feedback
• Customizable analysis prompts with persistent settings
• Cross-platform support (Windows, macOS, Linux)

🚀 Quick Start

Prerequisites

• Install Ollama and pull a model:

  # Install Ollama: https://ollama.ai/
  ollama pull llama3.1:8b
  ollama serve
  

• Use Chrome or Edge browser (required for Web Speech API)

Installation

# Install globally
npm install -g real-time-speech-analyzer

# Start the analyzer
speech-analyzer

Or run locally:

# Clone and run
git clone https://github.com/speech-analyzer/speech-analyzer
cd speech-analyzer
npm install
npm start

Usage

• Start the server - Opens at http://localhost:3000
• Configure analysis instructions - Default includes:
  • Sentiment analysis (5 keywords)
  • Controversial statement detection with counterarguments
• Click "Start Listening" and begin speaking
• View real-time results in the AI Analysis column

🎯 Default Analysis Instructions

• "Analyse sentiment, show as 5 keywords, nothing else."
• "Identify most controversial statement and respond with a counterargument."

🔧 Configuration Options

Analysis Instructions

• Add multiple instructions with the + Add button
• Update All - Apply changes to all fields
• Default All - Reset to default instruction set
• Remove fields with ✖ button (when multiple exist)

Context Management

• Summary threshold - Create summary after N chunks (default: 20)
• Recent chunks - Keep N recent chunks in context (default: 10)
• Force Summary - Manually trigger summarization

System Prompt

• Customize the AI's behavior and response style
• Default includes 25-word response limit
• Update and Default buttons for easy management

🖥️ Cross-Platform Support

Windows

npm install -g real-time-speech-analyzer
speech-analyzer

macOS/Linux

npm install -g real-time-speech-analyzer
speech-analyzer

Runtime Requirements

• Node.js 18+ (built-in support)
• Bun (optional, for faster performance)
• Ollama running locally
• Chrome/Edge browser

🏗️ Architecture

• Frontend: TypeScript + DOM manipulation
• Backend: Bun server with WebSocket communication
• LLM: Local Ollama integration
• Speech: Web Speech API with interim results
• Context: Rolling summary system for conversation memory

📊 Processing Flow

• Speech captured → Web Speech API
• Live transcription → Shows words as spoken
• Final chunks → Sent to analysis queue
• Processing notification → Yellow indicator with spinner
• Multiple LLM calls → One per analysis instruction
• Results displayed → Separate boxes per instruction
• Context updated → Rolling summary system

🔒 Privacy

• 100% local processing - No cloud APIs
• Your speech never leaves your machine
• Local Ollama models for AI analysis
• No data collection or tracking

⚙️ Development

# Development mode with hot reload
npm run dev

# Build for production
npm run build

# Start production build
npm start

Project Structure

speech-analyzer/
├── server.ts          # Bun WebSocket server
├── app.ts             # Frontend TypeScript
├── index.html         # Main UI
├── bin/               # Executable scripts
└── dist/              # Built files

🤝 Contributing

• Fork the repository
• Create a feature branch
• Make your changes
• Test cross-platform compatibility
• Submit a pull request

📝 License

MIT License - see LICENSE file for details

🐛 Troubleshooting

"Speech recognition not supported"

• Use Chrome or Edge browser
• Ensure HTTPS (or localhost)

"Ollama: Unavailable"

• Start Ollama: ollama serve
• Pull a model: ollama pull llama3.1:8b
• Check Ollama is running on port 11434

"No analysis appearing"

• Check browser console for errors
• Verify Ollama model is loaded
• Ensure analysis instructions are configured

Performance Issues

• Use lighter Ollama models (e.g., llama3.1:8b vs llama3.1:70b)
• Reduce context window settings
• Close other browser tabs

🔗 Links

• GitHub: https://github.com/speech-analyzer/speech-analyzer
• npm: https://npmjs.com/package/real-time-speech-analyzer
• Ollama: https://ollama.ai/
• Bun: https://bun.sh/

Made with ❤️ for real-time speech analysis