Generative Architecture

Overview

Generative Architecture

My role: Design and PoC development.

Challenge: Create randomised virtual properties within pre-defined rarities for an NFT metaverse environment.

Proof of Concept: Built a technology demonstrator in Unity 3D that procedurally generated landscape-aligned properties and props, respecting the rarity model.

Impact: Delivered a working prototype to validate feasibility, clarify requirements, and inform the engine/tooling decision (Unity vs Unreal; Blender for the creative pipeline).

Technologies: Unity 3D, C#

An R&D Excursion, with Transferable Skills

This project was intentionally outside the usual enterprise domain: it was a focused R&D excursion into procedural generation, 3D tooling, and rapid prototyping. The transferable value is the same as in enterprise delivery: turning ambiguous requirements into a working system, defining the right abstractions, and de-risking key technical decisions early.

Design Parameters and Constraints

As Metaverses proliferate, the demand for beautifully designed, unique virtual properties skyrocketed. But to make this an attractive proposition, the landscape and the properties have to fit seamlessly, indeed as they would in the real world.

When a collection of NFT properties is designed, there are several different main structures, which I call skeletons, and features like walls, roofs, doors, windows are added to them. Each feature may also have one or more randomisable parameters, e.g., wall or roof colour, window style etc. There are also props, which could be imagined as what goes in the garden around the house: flowers, trees, benches, fountains etc.

The challenge is that in a generated landscape, the position of the main structure and surrounding props need to be matched to where the landscape can properly accommodate them.

Unique Each Time

So the challenge is to:

  1. generate a landscape that can actually accommodate the desired number of properties
  2. randomise the parameters of the properties against the rarity requirements
  3. slot in the properties in the right place on the landscape
  4. map the features and props in the skeleton to the landscape

The proof of concept I have developed generated a new set of houses with trees as props on a landscape each time it was run. It was used to assess and refine the technical requirements as well as Unity as a platform. Based on the assessment, Unreal Engine was eventually chosen, due to visual capabilities required for the render pipeline, with Blender as the design platform for the creative team.