What Happened
The new web app at https://www.whodunnitai.com/ turns a murder mystery into a voice‑only game. Players record questions; the audio is sent to an AI model that replies as a suspect. Speech‑to‑text transcribes the input, the language model processes it, and text‑to‑speech returns the answer. The result is a conversation where players interrogate suspects, collect clues, and solve the case—all without typing.
The project launched on Hacker News with a “Show HN” post. Developers highlighted the novelty of mixing narrative mystery with real‑time voice interaction. They stress that the system uses open‑source components and can run locally or in the cloud, showing how voice‑enabled AI can scale.
Why This Matters for Builders
- Voice as a first‑class input channel: Speech can replace typed queries in complex workflows. Builders can design n8n or AI‑agent flows that accept voice commands, boosting accessibility and engagement.
- Real‑time transcription and response: The app shows a low‑latency pipeline from microphone to AI model to audio output. This pattern applies to production systems needing instant feedback, like support bots or interactive tutorials.
- Modular architecture: The project separates speech‑to‑text, natural‑language understanding, and text‑to‑speech. Teams can cherry‑pick these modules and plug them into existing workflows, saving time on core components.
- Data privacy and compliance: The open‑source design and local‑host option demonstrate how sensitive conversations can stay on premises. This is essential for regulated industries that must keep user interactions private.
- Narrative‑driven state management: The game tracks suspect responses and player progress with a structured state machine. Automation teams can adopt this approach to maintain context across multi‑turn interactions.
FAQ
Q: Can I use the same voice‑to‑AI pipeline in my n8n workflows?
A: Yes. n8n’s HTTP request nodes can call any speech‑to‑text or text‑to‑speech API, and you can invoke language models via REST or GraphQL. Combine these nodes to build a voice‑enabled workflow.
Q: What are the performance implications of adding voice to a production agent?
A: Speech processing adds latency, but efficient cloud services (e.g., Whisper or Google Speech API) can keep round‑trip times under a second. Caching common utterances further reduces delays.
Q: How do I handle user privacy when using third‑party speech APIs?
A: Review the provider’s data handling policies, consider on‑premises solutions, or use privacy‑preserving models that delete transcripts after processing. Always inform users how their voice data is used.
Originally published on Automations Cookbook.
Top comments (0)