All products

Assemble a PlayCanvas scene with AI

Place calibrated objects in a scene and check that their visible geometry, ground contact and collision bounds agree.

You can assemble a PlayCanvas scene with AI when individual assets look correct but do not fit together in the room. A chair can appear grounded from the front while its rear legs intersect a platform. The assemble-scene skill gives each object one placement root and keeps its calibrated visual beneath it, so position and collision remain associated.

The agent uses the object’s measured footprint rather than guessing a gap from the distance between pivots. Once the scene renders, another camera angle and a measurement at the actual point of contact can reveal a floating object or a genuine intersection.

What it does

  • Place several calibrated chairs around an existing table.
  • Keep collision shapes aligned with the visible object.
  • Give repeated objects predictable names for later edits.
  • Check whether overlapping bounds represent a real visual collision.
  • Remove an object along with the events its root owns.

Install and use

  1. Prepare an authorised PlayCanvas Engine project and suitable models. Have the physics runtime available when rigid bodies are involved.
  2. Ask the agent to place four chairs against the floor and verify their spacing from two views.
  3. Inspect the rendered scene and correct the model calibration first if its size or orientation is wrong.

FAQ

Source

Made by PlayCanvas. Original skill: assemble-scene, published on GitHub under the MIT licence.

View on GitHub