Calibrate PlayCanvas models with AI
Give each reusable GLB a consistent scale, orientation and support point so a second instance lands where its placement coordinates say it should.
Calibrate PlayCanvas models with AI when two objects placed at floor height look as if they belong to different worlds. A chair may float because its origin sits halfway up the back, while a shelf may face the wall because its source file uses another forward direction. The calibrate-model skill keeps the correction with the asset, ready for each new instance.
Choose one dimension that you can verify, such as a known shelf width or a door height. The agent can derive a uniform scale from the measured geometry, then set the visual offset and orientation independently. A placement point can now describe the shelf’s position in the room without carrying a new set of guesses.
What it does
- Scale a GLB against a known measurement or a previously calibrated object.
- Find the visual offset that makes the feet meet the floor.
- Check which way the asset faces in a running scene and store that rotation.
- Place a wall-mounted item using its mounting point.
- Inspect an animated character in motion when its initial pose gives misleading bounds.
Install and use
- Supply the PlayCanvas project, source GLB, a reference dimension and access to a geometry inspection tool and preview.
- Ask, “Make this shelf one metre wide, face it into the aisle and record the settings for the next shelf.”
- Place a second instance at a different support point and compare its size and contact with the first.
FAQ
Source
Made by PlayCanvas. Original skill: calibrate-model, published on GitHub under the MIT licence.
View on GitHub