Multimodal Spatial Web Infrastructure

Autonomous Pixel Avatar Synthesis & Interactive Event Village

Aeidos Labs powers Bangmyeonglok (방명록)—a real-time spatial guestbook bridging physical celebrations with top-down Hanok × Florida village environments. We combine on-device computer vision and Anthropic Claude APIs to orchestrate procedural pixel avatars, dynamic in-village advertisements, and multilingual community translations.

Live Product Demonstrations

📺 Launch Village (TV Demo) 📱 Mobile Guest Flow 🏛️ Entrance Kiosk 🔑 Admin Dashboard

Multimodal Vision-to-Avatar Pipeline

Claude 3.5 Sonnet Vision

Transforms attendee selfies into custom 22×34 pixel avatars with procedural hair, facial structures, skin tone lighting calibration, and attire. Upgrading from client-side heuristics to Claude Vision ensures high-fidelity tokenized feature extraction under challenging event lighting.

Contextual In-Village Billboard Matching

Claude 3.7 / 3.5 Sonnet

Performs real-time semantic analysis over guestbook logs and sentiment markers. Dynamically matches event sponsors and synthesizes high-converting, personalized billboard greetings displayed across virtual town plazas without human intervention.

Multilingual Streaming & Event Narrative Engine

Claude Streaming API

Provides sub-second contextual translation preserving celebratory nuances across Korean, English, and global attendees. Post-event, synthesizes hundreds of messages, audio transcripts, and time capsules into automated Instagram-ready (1080×1350) narrative reports.

Spatial IoT Sync & Low-Latency Remote Controls

Edge Architecture

Bridges personal mobile remotes (hearts, claps, dancing, bouquet toss) directly to shared TV display canvases. Features in-memory Firestore caching, sub-600KB payload compression, and smart neighbor rotation to protect cloud resource quotas.

Claude 3.7 Sonnet Claude Vision HTML5 Canvas (22x34) Firebase Firestore Cloudflare Pages

Contact: contact@bangmyeonglok.com