A voice-powered murder mystery game where players interrogate AI suspects through natural conversation to solve complex cases. Built to explore the intersection of conversational AI, real-time voice processing, and interactive storytelling.
Players use voice commands to question AI-driven suspects, each with distinct personalities, backstories, and secrets. The game combines speech recognition, natural language processing, and text-to-speech to create immersive interrogation experiences that adapt to player questioning styles.
Real-Time Voice Processing
- Bidirectional voice communication with sub-200ms latency
- Voice activity detection and conversation turn management
- Multi-speaker audio processing and isolation
AI-Driven Characters
- Dynamic personality modeling with persistent memory systems
- Context-aware response generation maintaining character consistency
- Adaptive dialogue that responds to investigation patterns
Game Logic
- Evidence tracking and deduction scoring algorithms
- Multi-path narrative branching based on player choices
- Real-time accusation validation and case resolution
Frontend: Next.js with TypeScript, real-time audio via LiveKit SDK Voice Pipeline: Deepgram (STT) → OpenAI (LLM) → Cartesia (TTS) Infrastructure: Python-based agent system with WebRTC for low-latency communication
Node.js 18+, Python 3.9+, pnpm
API keys: LiveKit, OpenAI, Deepgram, Cartesia# Frontend
cd frontend && pnpm install
# Add LiveKit credentials to .env.local
pnpm dev
# Agent
cd agent && python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt
python3 agent.py download-files
# Add API credentials to .env.local
python3 agent.py devAccess at http://localhost:3000
Built to explore voice AI applications in interactive media and test real-time conversation systems at scale.