On the workbench
Proposed experimentOne lantern, three paths into a game
What changes when three workflows make the same stylized lantern for the same game scene?

The comparison
Build the same stylized lantern by hand, with AI-assisted concept development followed by a hand-built mesh, and with a generated mesh followed by cleanup. Put each version in the same small engine scene with the same camera, lighting, target size, and interaction.
What we’ll look for
Compare silhouette at play size, repair effort, import, material and lighting control, and how each asset handles a separate collider. Record the tools, settings, rights, and actual time spent once the experiment is run.
For the ideas behind this comparison, read When a beautiful prop is not yet a game asset.