Hello: Earth

Our agent turns a browser window into a gesture-driven travel guide: you raise your hand, spin a 3D globe, “hand-pick” a city, and a conversational guide speaks back with fun, structured facts, local flavor, and simple follow-up questions. The working prototype focuses on a learning path: hand detected → city locked → first answer → one follow-up—backed by pre-baked fallback content so the demo is stable even if the LLM or network hiccups. Technically, it orchestrates multiple modalities and tools: webcam + gesture tracking, a Three.js 3D globe for spatial UI, a cloud LLM (e.g. Gemini 2.0 / GPT-4.x) for city intelligence, and ElevenLabs real-time TTS API for low-latency voice, all coordinated via a lightweight browser state machine. Innovation & Creativity come from treating “where do you want to go?” as an embodied, playful interaction instead of a dropdown—your hand becomes the cursor for the planet, and the agent turns micro-facts into tiny stories that make distant places feel familiar. Real-World Impact shows up in obvious extensions: conference wayfinding, classroom geography, global hackathons, and museum-style exhibits where people learn about hubs, partners, or cultures through a single, approachable interface.

For Theme Alignment, the project literally fuses browser (WebGL/Three.js front end), voice (ElevenLabs streaming TTS), cloud (LLM inference via Gemini/GPT), and tools (webcam + hand-tracking + simple UI state management) into one cohesive agent loop: vision input from the camera becomes an intent (selected city), which drives a cloud LLM call, which is rendered as both on-screen text and real-time speech.