Course 2

Game Asset Optimization with Agentic AI

Take the same models and make them game-ready: low-poly retopology, UV unwrapping, and baked maps.

The optimization stage is where agents are strong but quietly lossy. Learn to drive budgets, UVs, and bakes — and to measure what was lost.

Choose your track

Track A (designers/artists): focus on judgment — reading agent output critically and steering it toward your intent.

Beginner

First prompts, basic constraints, reading agent output.

Your first retopo prompt and the triangle budget

ABC ~7 min

Tell the agent to "make it low-poly" and it will either decimate the model into a crumpled silhouette or shave off two hundred triangles and call it optimised — "low" is a direction, not a target, and the agent has no idea which one you meant.

Intermediate

Multi-step prompts, references, recovering from common failures.

UVs the agent quietly gets wrong

ABC ~7 min

The agent will report "UV unwrap complete" over a layout with overlapping islands, seams running across visible faces, and texel density that swings 4x across the model — none of which shows up until a texture is applied and the seams scream.

Advanced

Repeatable workflows, quality gates, edge cases the agent fumbles.

Baking high-poly detail onto the low-poly

ABC ~8 min

Baking is where the agent confidently produces a normal map full of skirts, waves, and black gaps — because it set a cage and ray distance it cannot see the consequences of, and a render thumbnail will not reveal the artifacts until they ship.

Master

The capability boundary: where you stop trusting the agent and take over.

Passing the asset cleanly into an engine

ABC ~7 min

The export is where everything you got right quietly breaks: the agent exports a glTF that loads into the engine at 1/100th scale, with flipped normals and the normal map unhooked — each a one-line setting it never thought to check.

← All three courses