Insights · guide
Test procedural placement with cubes before real assets
Drafted with Insight · edited by The WE ARE VR team
development

A grey cube is often the fastest route to a believable world.
I wrote this because procedural placement can waste days before anyone notices the rules are wrong. For Jozi teams building activations, demos, or commissioned XR work, that delay reaches artists, producers, and venue crews quickly.
What's the problem?
Procedural placement means asking code to position objects automatically. You might scatter trees, street furniture, products, enemies, or training hazards across a scene.
The first version often uses finished models. That feels productive because the scene becomes attractive early. It also hides basic faults.
A detailed tree can disguise a bad pivot point, which is the model's placement origin. A wide product display may overlap its neighbour while its visible shape makes the collision hard to read. Large textures and complex geometry can slow the headset, adding performance noise to a placement test.
Then feedback gets muddled. The artist adjusts a model. The developer changes the spacing. The producer asks why objects still appear inside walls. Everyone is solving a different problem.
In a Johannesburg activation, you may also be testing inside a tight booth with fixed play boundaries. A spawn error near the virtual edge can become a real safety problem when a guest reaches towards it.
What's the solution?
Use one plain cube as a stand-in for every placed object.
Give the cube the same approximate width, depth, and height as the future asset. Add the same collision shape you expect to use later. A collider is the invisible boundary the engine uses to detect contact and overlap.
Now test four rules separately:
- Where may an object spawn?
- How far must it sit from another object?
- Which surfaces or volumes are forbidden?
- What happens when no valid position remains?
Colour can carry one extra piece of information. Green cubes passed every check. Red cubes failed a spacing or collision check. Yellow cubes used the final allowed retry. Keep the palette simple (the cube is already doing unpaid overtime).
Imagine a client commissions an XR retail layout for a Sandton event. Ten product plinths must appear inside a six-by-four-metre virtual zone. Each needs one metre of clear walking space, and none may block the entrance.
Start with ten cubes sized like the plinths. Draw the allowed zone on the floor. Add a forbidden box at the entrance. Run the placement rule fifty times with different random seeds. A random seed is a saved starting value that lets you reproduce the same arrangement later.
If cubes overlap, bunch near one wall, or fail after seven placements, the problem belongs to the rule. You can fix it before finished models arrive. Once the cube test passes consistently, swap in real assets while keeping the proven positions, dimensions, and collision checks.
Record one passing seed and one failing seed. The failing case is especially useful because it gives the team a repeatable bug instead of a vague screenshot.
What's next?
- Replace one procedurally placed asset with a cube matching its rough dimensions.
- Run at least twenty seeds, checking spacing, walls, doors, and player boundaries.
- Save screenshots and seed values for one pass and one failure, then share them with the artist.
Share your cube test, oddest spawn failure, or placement question with the Jozi XR community.
The WE ARE VR team
