Skip to main content

Mesh instancing

Instancing draws many placements of shared geometry with one draw per compatible primitive instead of duplicating models.

The SimpleInstancing sample demonstrates this capability. Drag to orbit and scroll to zoom.

GitHub
Loading examplePreparing GPU resources…

API​

Engine Model supports instanced attributes and an instanceCount. The glTF adapter creates those instanced models automatically from authored transforms. getGLTFNodeInstancing() exposes resolved matrices and source attributes for inspection or a custom integration.

import {createScenegraphsFromGLTF, getGLTFNodeInstancing} from '@luma.gl/gltf';

const scenegraphs = createScenegraphsFromGLTF(device, gltf);
for (const sourceNode of gltf.nodes || []) {
  const instancing = getGLTFNodeInstancing(gltf, sourceNode);
  if (instancing) {
    console.log(instancing.matrices.length, instancing.attributes);
  }
}

Authored instances share render state. Independent skeletal animation uses the animated crowd API rather than ordinary static instancing.

glTF support​

EXT_mesh_gpu_instancing supplies accessor-backed transforms and custom _NAME attributes. Each source primitive becomes one instanced model on WebGPU and WebGL. See native glTF extensions.

See glTF materials and textures, glTF animation, and the extension support matrix for adapter details and current support limits.

All engine features