Offline AI Space Guide (Voice-Driven VR)
The Offline AI Space Guide is a voice-driven VR experience powered by three local AI models running simultaneously on-device — speech recognition, a large language model, and text-to-speech — with zero cloud dependency.
Speak naturally. The guide listens, thinks, and responds in real time, entirely offline.
Why it matters:
Most AI-powered VR experiences require a live internet connection. This project explores what becomes possible when the full AI pipeline lives on the device — enabling use cases in remote training, data-sensitive environments, and regions with unreliable connectivity.
Technical breakdown:
ASR: Recognissimo (on-device speech recognition)
LLM: Llama via UndreamAI (local language model)
TTS: Lingotion Thespeon (on-device voice synthesis)
Built in Unity Engine with C#