Project Skills

Skills and technologies required for each Metason deployment project, derived from our company profile and case portfolio.

City-scale Dynamic Digital Twin of Hong Kong

-Petabyte-scale digital twin of Hong Kong built on Common Spatial Data Infrastructure (CSDI) datasets and schema from the HKSAR Government.
-3D Map, 3D Pedestrian Network, point cloud, BIM and 360 imagery are accessible in one Omniverse desktop application.
-City model in OpenUSD with Postgres/PostGIS and NVIDIA Nucleus — a visible, queryable twin rather than a static 3D viewer.

Live city-scale digital twin of Hong Kong on NVIDIA Omniverse

LIVE city twin

Project: Dynamic Digital Twin of Hong Kong

Urban canyon view inside the Hong Kong digital twin

City-scale OpenUSD scene

Project: Dynamic Digital Twin of Hong Kong

System architecture

Project: Dynamic Digital Twin of Hong Kong

NVIDIA Omniverse & OpenUSD

-NVIDIA Omniverse is the deployment foundation for photorealistic simulation and interoperability across 3D, GIS and AI tools.
-CSDI and BIM sources are converted to Omniverse-ready OpenUSD for USD Composer, Presenter and Viewer workflows.
-NVIDIA Inception company — GTC 2024 showcase and official NVIDIA solution brief for the Hong Kong city twin.

GTC Dynamic Digital Twin applications converting multiple datasets to OpenUSD

GTC DDT Applications

Project: NVIDIA GTC 2024

Hong Kong city twin running in Omniverse with RTX real-time viewport

Omniverse RTX viewport

Project: Dynamic Digital Twin of Hong Kong

CIC 2024 Omniverse demo

Project: CIC 2024

CSDI data processing & BIM → CityGML → USD

-Automated ingestion, cleaning and validation of CSDI-aligned geospatial data — more than 200 TB, ~1 million files and ~1,300 datasets.
-BIM (Revit, Civil 3D, IFC) converted to CityGML and OpenUSD with Postgres/PostGIS spatial search for Lands Department SMO.
-2D GIS layers translated into 3D spawnable assets: point cloud, FBX city models, 360 imagery and 3D pedestrian routes.

Lands Department BIM to CityGML spatial database pipeline

BIM → CityGML → USD

Project: Lands Department SMO

Geospatial data processing for the industrial metaverse

Data processing pipeline

Project: CSDI / OpenUSD conversion

Natural language to SQL over CSDI geospatial databases

Text-to-SQL on CSDI

Project: iGeoCom / CSDI

Spatial LLM, RAG & tool calling

-LLM agents inside Omniverse call unique tools: spawn 3D models, shortest path, isochrone / living-circle analysis, and persona-based layer filters.
-RAG on Hong Kong land law and property knowledge, plus Text-to-SQL over CSDI / PostGIS so planners can ask the twin in natural language.
-Urban planner, surveyor and general-user personas — OZP / BHC, survey grid spawn, and address-based navigation.

AI chat with the digital twin for urban planning queries

Urban planning chat

Project: URA / Omniverse AI Agent

RAGFlow agent translating natural language into CSDI SQL

RAG + Text-to-SQL

Project: HK CSDI

Tool calling in Omniverse

Project: Dynamic Digital Twin of Hong Kong

Agentic AI — skills & memory

-From fixed tool calls to agents that plan, use skills and remember — a self-improving digital twin loop (planner / skills / memory).
-Case study: 5-minute living circle — the agent searches CSDI data, highlights results in the twin, and exports a planning PDF.
-Cross-application agents connect Omniverse with design tools, government datasets, flight-path planning and drone simulation.

Agent spawning MegaBox and surrounding models for living-circle analysis

5-minute living circle

Project: Agentic Digital Twin 2026

15-minute life circle analysis report generated from the twin

Agent-generated planning report

Project: Agentic Digital Twin 2026

Agent task run

Project: Agentic Digital Twin 2026

Simulation on the Twin — water, wind & Isaac Sim

-The same Omniverse twin hosts physics simulation: water behaviour, city-scale air ventilation CFD, and robot training.
-Collaboration with WSP — Ansys Fluent CFD visualised as live streamlines on Hong Kong 3D Map models.
-Isaac Sim on the Hospital Authority campus twin for multi-AMR missions, occupancy maps and trajectory training.

Water simulation

Project: WSP × Omniverse 2024

Wind / air ventilation CFD

Project: WSP × Omniverse 2025

Isaac Sim robot training on a hospital campus digital twin

Isaac Sim AMR training

Project: Hospital Authority

Blender: 3D modeling & animation

-Modeling, sculpting, rigging, animation and rendering in Blender.

Rigging

Animation

Project: Virtual Artist "Vela"

Hair modeling & simulation

-Blender's Hair Particle System: generate and style strand-like objects such as hair, fur, and grass.
-Hair Curves Dynamics: add physics directly to your hair curves. In the Physics Properties tab, enable Simulation and adjust settings like Stiffness, Mass, and Damping.

Hair preview

Project: Virtual Artist "Vela"

Particle System

Hair Simulation

Project: Virtual Artist "Vela"

Video editing

-Perform basic actions like video cuts and splicing, as well as more complex tasks like video masking or color grading.

Video cuts

Project: Virtual Artist "Vela"

Color grading

Video masking

Marvelous Designer

-It features a seam-and-sewing workflow that easily stitches fabric over rigged 3D character models.

Project: Virtual Artist "Vela"

Project: Mina

Project: Mina

Unreal engine: MetaHuman

-Create a realistic human 3D model.

MetaHuman

ComfyUI: Text to images (Z-Image Turbo)

-A text-to-image AI model developed by Alibaba's Tongyi Lab. It produces photorealistic images in under a second using only 8 steps. It runs locally on consumer GPUs with 16GB VRAM and supports precise English and Chinese text rendering.

Z-Image Turbo workflow

Text to image

Text to image with LoRA

Project: Virtual Artist "Vela"

Image edit (Qwen 2.5 11)

-An advanced open-weights image editing model by Alibaba, excelling in instruction-based image manipulation, text rendering, and character consistency.