Persona.AI - Master Your Communication Skills

An advanced AI-powered communication coaching platform that provides real-time feedback on presentation skills using multi-modal analysis.

Features

• Real-time AI Analysis: Advanced computer vision and audio analysis using MediaPipe and TensorFlow.js
• Multi-modal Feedback: Analyzes facial expressions, posture, voice quality, speech clarity, and content engagement
• Category-specific Coaching: Tailored analysis for job seekers, professionals, speakers, sales teams, remote workers, and students
• Privacy-first: All processing happens locally in the browser with optional backend analysis
• Comprehensive Metrics: Eye contact, posture, gesture analysis, speech patterns, and emotional recognition

Technologies Used

• Frontend: React 18, TypeScript, Vite
• UI Framework: ShadCN/UI, Tailwind CSS, Radix UI
• AI/ML: MediaPipe (Face Mesh, Pose, Gesture), TensorFlow.js, Hugging Face Transformers
• Audio Processing: Web Audio API, YIN Pitch Detection, SNR Analysis
• Backend: Supabase (optional for extended analysis)
• State Management: TanStack Query, React Router

Project Structure

src/
├── components/          # Reusable UI components
├── pages/              # Main application pages
├── lib/                # Core analysis algorithms
│   ├── visionAnalysis.ts    # MediaPipe-based vision analysis
│   ├── audioAnalysis.ts     # Audio processing and analysis
│   ├── speechRecognition.ts # Speech-to-text and analysis
│   ├── contentAnalysis.ts   # NLP content analysis
│   └── fusionAlgorithm.ts   # Multi-modal metric fusion
├── hooks/              # Custom React hooks
├── integrations/       # External service integrations
└── types/              # TypeScript type definitions

Key Components

Vision Analysis

• Face Detection: MediaPipe Face Mesh (468 landmarks)
• Emotion Recognition: FACS-based emotion detection with blendshapes
• Eye Contact Tracking: Iris position analysis with gaze vector calculation
• Posture Analysis: 3D pose estimation with joint angle analysis

Audio Analysis

• Pitch Detection: YIN algorithm for fundamental frequency
• Voice Quality: SNR (Signal-to-Noise Ratio) for clarity
• Speech Patterns: Filler word detection, WPM calculation

Content Analysis

• NLP Processing: TF-IDF keyword extraction, sentiment analysis
• Readability Metrics: Sentence complexity and vocabulary richness
• Named Entity Recognition: Pattern-based entity extraction

License

This project is licensed under the MIT License.