Show HN: Voice driven murder mystery, Interview AI suspects with your voice
This "Show HN" introduces WhoDunnitAI, a novel voice-driven murder mystery game where players interrogate AI suspects using real-time speech-to-speech technology. The project leverages advanced OpenAI models for both suspect interaction and evidence validation, showcasing a creative application of current AI capabilities. Its blend of interactive gameplay and cutting-edge AI implementation makes it a compelling submission for the Hacker News community.
The Lowdown
WhoDunnitAI presents an innovative murder mystery experience where players step into the role of a detective, interviewing AI-powered suspects using their own voice. The project, initially a proof-of-concept, has been revitalized to take advantage of more mature AI voice agents, offering a unique interactive narrative.
- Voice-Driven Interrogation: Players engage in speech-to-speech interviews with AI suspects, powered by OpenAI's
gpt-realtime-2.1model over WebRTC. - Evidence Validation System: When players make an accusation, suspects log the claim and presented evidence, which is then judged by a
gpt-5-minimodel to verify if genuine case facts were articulated. - Technical Stack: The application is built using Next.js, MongoDB, and Clerk for authentication.
- Resource Management: Due to the high cost of the
gpt-realtime-2.1model, the experience includes a 30-minute time limit per session and requires authenticated user IDs to manage expenses.
This project stands out as a creative and technically ambitious use of AI to deliver an engaging, interactive storytelling experience, pushing the boundaries of what's possible in digital games.