Modeling I: topology & the polygon budget

Mon 2:00 · Brief The mesh is a budget

Last week you saw where art sits in the pipeline. This week you learn the first hard constraint: every vertex costs something, and the GPU is counting.

Quads, tris, and what actually ships

You model in quads because quads subdivide predictably, deform cleanly when rigged, and give you readable edge flow. But the GPU doesn't render quads — at draw time everything is triangulated. A quad is two triangles; an n-gon is however many the triangulator decides, and it may decide badly, producing shading errors you didn't author. So: author in quads, think in triangles. The triangle is the unit you actually spend.

Turn on the Statistics overlay (Overlays → Statistics) and keep the tri count visible while you work. That number is your bank balance. A prop, a wall segment, and a hero character all have wildly different budgets, and the number tells you the truth your viewport lighting is hiding.

Edge flow, poles, and n-gons

Edge flow is the direction your loops travel across a surface. Good flow follows form and keeps faces close to rectangular; bad flow pinches, twists, and creates poles — vertices where three or five edges meet instead of four. Poles aren't forbidden, but they concentrate shading and deformation problems, so you place them deliberately, in flat or hidden areas. N-gons (faces with more than four sides) are the other trap: fine as scaffolding while you block, unacceptable in a final game mesh because you've handed shading decisions to the triangulator.

Modular kits and the grid

Professional environments are rarely one giant model. They're kits — a small set of pieces (wall, corner, doorway, trim) built to a shared grid so they snap together in any combination. Design to the grid from the first vertex: pick a unit (say 2 meters), enable snapping, and make sure your panel's edges land exactly on grid lines so copies tile seamlessly. A kit that snaps is a kit that scales; a kit that's a few centimeters off is a kit of gaps.

Mon 2:45 · Guided lab Block a modular wall panel

One panel, done right, on a grid. We build the discipline before we build the kit.

1. Set your unit and grid

   Decide a module size — 2m wide, 3m tall is a good wall. Confirm your scene units and turn on snapping (Snap → Increment, magnet icon). Every piece in the kit will share this grid, so commit to it now.

2. Block the silhouette first

   Start from a plane or cube scaled to your module and rough in the big shapes — the panel's profile, a recessed section, a raised rib. Silhouette reads before detail does; get the outline right before you add a single bevel.

3. Keep it in quads

   Use loop cuts (Ctrl+R) and inset (I) to add definition, and watch that faces stay four-sided. If you create an n-gon while blocking, note it — you'll resolve it before this is a final mesh.

4. Control the edge flow

   Route loops so they follow the panel's forms and stay rectangular. Push any unavoidable poles into flat interior areas where shading won't reveal them. Rotate around the model and check that loops read cleanly from every angle.

5. Watch the tri count

   With the Statistics overlay on, glance at your triangle total after each major addition. Set yourself a budget for a single wall segment and stay under it — a wall is not where your polygons should go.

6. Make it tile and snap

   Verify the panel's left and right edges sit exactly on grid lines and at matching heights, then duplicate it (Shift+D) and slide the copy one module over. The seam should vanish. If it doesn't, your edges aren't on the grid — fix the original, not the copy.

7. Name and organize

   Rename the object and its mesh data to something meaningful like SM_Wall_Panel_A, apply transforms (Ctrl+A → All Transforms) so scale and rotation are clean, and set the origin sensibly for grid placement. Clean naming now saves you an hour in Unreal later.

Wed 2:00 · Workbench Practice, unhurried

No new material. You work on your own file while I circulate — this is the hour where the technique from the guided lab stops being something you watched and becomes something you can do.

1. Finish the lab on your own work

   The guided lab was built on the class example. Now do it on the piece you are actually submitting. That second pass is where it sticks.

2. If you fell behind, catch up here

   This is what the hour is for. Say so and I will come to you — do not spend it quietly stuck.

3. If you are ahead, go deeper

   Take the optional extensions from the lab, or push the piece further than the brief asks. Nothing graded depends on them.

4. Ask Gary, the AI TA, anything

   The tutor on the course site is open all term for questions about how and why. It explains; you build. No agents in the work until Week 9.

Wed 3:10 · Crit / share Wireframe on the wall

Everyone drops their panel into wireframe view and we look at edge flow together on the projector. We're not judging how sci-fi it looks yet — we're reading topology: are the faces quads, does the flow follow the form, where did the poles land, and does the tri count match the budget the panel deserves? Bring the kit spec your agent wrote; we'll compare how different budgets led to different-looking panels for the same brief.

Homework — due before Week 03

Resources

• Blender mesh modeling docs — loops, poles, and clean topology tools.
• Unreal Nanite docs — why budgets still matter even with virtualized geometry.
• Unreal mesh import docs — what happens to your geometry on the way in.