From the physical world or a reference to usable 3D geometry, and generation or modification of geometry from natural language. Capture through neural rendering, Gaussian Splatting and similar techniques, yields a model that integrates directly into the design flow without the heavy post-processing of classical photogrammetry. Language-assisted modelling lets the architect describe what is needed, from adjusting an envelope to the constraints of the site to generating façade-system variations or repairing imported geometry, and the AI produces the executable script.
F3
Module F3 of 10
AI Modeling
From concept to controlled geometry — at design speed
4 ECTSOne fortnight6 sessionsOnline, live and recorded
The module
Gallery
Objectives
- Capture real space as workable geometry.
- Understand the difference between capture, mesh and parametric model.
- Drive modelling operations from natural language.
- Repair and prepare geometry for downstream use.
What you will be able to do
Learning outcomes
The student integrates a capture model of a real space into their design environment and performs language-driven modelling operations on their own project.
Professional outcomes
Removes the bottleneck between survey and model, and lets small studios take on capture work that used to require specialists.
Methodology
Learn by doing, on the student's own real project. No software tutorials: a pre-built starting package is provided and base resources are given apart. Sessions combine a short input, guided practice and critique.
Concepts
Gaussian Splatting
Session by session
- Capture: from space to geometry
- Neural rendering and splatting
- Cleaning and repairing captured geometry
- Modelling from natural language
- Generating façade and envelope variations
- Integrating capture into the design flow
Deliverable
Deliverable F3: A capture model of a real space integrated into the design environment, plus a natural-language modelling exercise on the student's own project.
Capture is usable, not just presentable.
Geometry is clean and tolerant.
Language-driven operations are reproducible.
Integration with the design flow works.
Faculty and resources
Module Faculty
Learn this in MIAWS
This module is part of the 18-week postgraduate program. Work with your own real project from week 1. First cohort February 2027.
Talk to MIAWS AI →