The module works the step from the physical world or a reference to useful 3D geometry, and how to generate or modify geometry from natural language.
The first part covers capture: 3D reconstruction from video or images with neural rendering techniques (Gaussian Splatting and similar). The result is a model directly integrable into the design workflow without the intensive post-processing of traditional photogrammetry.
The second part covers language-assisted modeling: the architect describes what they need (adjusting an envelope to site constraints, generating variations of a facade system, debugging imported geometry with errors) and the AI generates the executable script in Grasshopper or Python.
Deliverable F2: 3D capture model of a real space integrated in the design environment, and a natural language-assisted modeling exercise applied to an element of the student's project.