Deliverable #22 · 2026-07-25
Sourcing 3D objects — why primitives, not sprite sheets
Backdated: added to the log on 2026-09-13 for a thing made on 2026-07-25. The original is docs/historic/3d-object-sourcing.html; what is embedded is a verbatim copy.
What it was
Kris had written into the recipe draft that 3D models would be easier than hand-drawn sprite sheets for a game where players choose their own body, colour and gear — and flagged it, in capitals, as an assumption to check. This note is the check. It tried three ways of sourcing models rather than reasoning about them: composing everything from Godot’s own primitives — boxes, cylinders, spheres, capsules, with flat colours and named attachment points; buying or downloading ready-made low-poly packs; and generating meshes. Each was scored against the mechanic the game needs, a horse plus a filled wings slot becoming a pegasus, a hat snapping to a socket on whatever head is there, and against iteration speed and cost.
What carried through
The verdict: primitive composition as the backbone. The assumption was right, and the reasoning is why. Every model in the game today is built that way — the creatures and gear of the object gallery first, and since then the Dirt clump, the fire and ice crystals, the stone golem, the Lightning Tower, the Grove Tree, the spell scroll, the glove worn by whoever is it in tag. None of them was bought or drawn; each is a small scene of shapes with sockets, which is what lets a hat fit any head and a horse become a centaur.
What did not
The morphing half of the note — continuous blends between shapes — has not been needed yet; the game’s models are discrete. And the question the note leaves open is still open: whether a creature’s shape and colour will one day be derived from its element data rather than placed by hand.
Sources & artifacts
← related epic: object-system-experiments
Also available as raw markdown.