VLAM!

A WebGPU Gaussian Splat viewer for three.js.

Docs and demo: https://vlam.voluma.ai · get started · guides

Local site and generated API: npm run dev http://localhost:5170

Warning

This is a pre-release not yet recommended for production. The API can still change in breaking ways before v1.0.

Install

npm install @voluma/vlam three

three is a peer dependency (>= 0.186.0). See the migration guide for 0.x API changes.

Usage

import * as THREE from 'three/webgpu';
import { SplatMesh } from '@voluma/vlam';
import { loadSplatData } from '@voluma/vlam/loaders';

const renderer = new THREE.WebGPURenderer({ antialias: true });
renderer.setSize(innerWidth, innerHeight);
document.body.appendChild(renderer.domElement);

const scene = new THREE.Scene();
const camera = new THREE.PerspectiveCamera(60, innerWidth / innerHeight, 0.01, 100);
camera.position.set(1, 0.5, 1.4);
camera.lookAt(0, 0, 0);

const splats = new SplatMesh(await loadSplatData('/scene.sog'));
scene.add(splats);

renderer.setAnimationLoop(() => {
 splats.update(camera, renderer);
 renderer.render(scene, camera);
});

SplatMesh works with a regular three.js WebGPURenderer (including its automatic WebGL2 fallback). For large streamed or unified scenes, createWebGPURenderer() is an optional convenience that requests higher WebGPU device limits and preserves MSAA where supported.

Supported formats: .sog, .ply, .spz, .splat, .ksplat, .lcc, .lcc2, .rad. Optional proxy-mesh relighting is available from @voluma/vlam/relighting; see the relighting guide. Support and remaining device-validation gates are tracked separately in the capability matrix and roadmap.

Contributing

See CONTRIBUTING.md and docs/architecture.md. By participating you agree to the Code of Conduct. Security reports: see SECURITY.md.

License

MIT. Third-party notices: THIRD_PARTY_NOTICES.md.

Made by Voluma.