Hard-Surface Vertex Normal Workflow Guide

1. Establish the modeling order

Start with real dimensions, transform scale, silhouette, and wall thickness. Resolve Boolean intersections before beveling so duplicate faces do not produce unstable highlights. Set bevel segments and radius from the camera distance, then add support loops or creases only where the subdivision result requires them. Delay custom normal work until these geometric decisions stop changing; every later topology edit can invalidate loop-normal data.

2. Choose smoothing boundaries

A hard edge splits the shading direction across an edge; a soft edge shares a smooth transition. In engines that split vertices at hard boundaries, align important hard edges with UV seams when practical. Smoothing groups express the same ownership decision in many packages. Smooth shading should make intentionally curved bands continuous, not round every planar panel. Preview with a glossy neutral material and moving light, not a busy texture.

3. Weight and edit vertex normals

Weighted normals bias each vertex normal toward larger faces, preserving broad flat panels while bevels catch light. In Blender, apply scale, set Shade Smooth or angle-based smoothing, configure the Weighted Normal Modifier after Bevel, and use Keep Sharp where marked boundaries must survive. A general normal modifier or direct normal edit is appropriate only for a documented target, because arbitrary edits are hard to reproduce after topology changes.

4. Control subdivision and topology

Subdivision Surface needs evenly distributed support around curved areas and deliberate poles away from highlight paths. Quadrangulate when deformation or subdivision continuity benefits, but do not convert every stable planar N-gon mechanically. Triangulate for final export when the target importer may choose a different diagonal. Inspect the triangulated result because a long diagonal across a non-planar face can create a visible shading fold.

5. Constraints and handoff settings

Freeze or apply transforms before evaluating bevel width. Record smoothing angle, modifier order, crease values, and whether custom split normals are exported. Avoid zero-width bevels, coincident intersections, and paper-thin shells. If the destination recalculates normals, either disable that import option or agree on the same angle and tangent basis. Retain a clean pre-triangulated source while shipping a deterministic runtime mesh.

6. Decision example: machined housing

A machined housing has large planar covers, a rounded grip, and narrow chamfers. Use hard boundaries around detachable covers, smooth the grip bands, and weight normals toward the covers after beveling. Keep a planar access-panel N-gon only if its triangulation is stable and no subdivision crosses it. Add support edges around the grip, verify wall thickness at vents, then lock triangulation before the normal-map bake.

7. Diagnose shading failures

A wavy panel usually points to non-planar faces, poor triangulation, or small bevel faces pulling shared normals. A dark seam can mean an unintended hard edge, mismatched smoothing, or duplicated vertices. Pinching under subdivision often comes from crowded support loops or a pole in the highlight path. If shading changes after export, compare modifier application, custom-normal export, importer recalculation, and final triangle indices in that order.

8. Evidence checklist and multilingual workflow table

Approval evidence should include a wireframe, glossy-light turntable, face-orientation view, triangulated preview, and destination-engine screenshot. Capture modifier order and smoothing settings. The table keeps review language unambiguous across teams; each row is a production decision, not an isolated vocabulary entry.

Related production references