
Generating and rigging game-ready 3D units for a web RTS
How we turn a text brief into an animated 3D unit for a browser RTS: Gemini concepts, Meshy meshes, agent-driven Blender rigging, and Mixamo clips, now open source.
Tag
Development posts tagged generative-art. Tags cut across the categories, so this pulls together work on the same subject from different parts of the game.
10 posts

How we turn a text brief into an animated 3D unit for a browser RTS: Gemini concepts, Meshy meshes, agent-driven Blender rigging, and Mixamo clips, now open source.

Fences, ruins, barricades and camp dressing are functions that build a mesh, not downloaded models. One painted sheet gives them all a hand-painted finish.

Every faction's walls are procedural 3D models in three upgrade tiers, drawn to match its buildings, with gates that open for friends and a pathfinder that knows.

A waygate ships as one unpowered picture. The light turning inside its arch is drawn by code, from a single field the 2D bake and the 3D shader both read.

Ninety generated letters, traced to vector and shipped as drawable paths. Split the brief into topology and execution, and keep fill out of the skeleton.

An eighty second trailer cut from eighteen generated clips, each one seeded with the game's own splash art so the dwarves stayed the game's dwarves.

Every building is generated twice: as a painted sprite, then as a 3D model built from that sprite. No human drew either one, or wrote the brief behind it.

Every axe landing, every worker's grunt and every unit that answers when you click it. All of it written out as prose first, then generated from the sentence.

Every unit and building in this game began as a generated image on a magenta field. What turns one into a sprite is five passes of unglamorous pixel work.

Wood elves fight to a fiddle reel at 145 BPM and dark elves to symphonic metal at 120. Nine factions, 171 tracks, and the difference written down first.