Skip to main content

@luma.gl/engine/splats

@luma.gl/engine/splats provides experimental GPU-native Gaussian splat rendering. It owns prepared splat data, directional spherical harmonics, semantic filtering, GPU picking, hierarchy traversal, bounded residency, decoded Khronos glTF attributes, covariance projection, depth ordering, and render models without depending on Apache Arrow, loaders.gl, glTF packages, or deck.gl.

The module is public with experimental APIs. Install it with yarn add @luma.gl/engine.

Interactive Gaussian splat showcase​

Explore a deterministic generated Gaussian scene without downloading a third-party capture. WebGPU projects, orders, and draws each progressively added batch through a GPU command graph.

GitHub
0s

Open the full Gaussian splat showcase.

Choose a topic​

TopicUse it for
RenderersPrepared, command-graph, segmented, and paged rendering contracts.
Data and shadingSource columns, covariance, spherical harmonics, semantics, and updates.
Picking and scenesGPU picking and mixed mesh-and-splat scenes.
Streaming and residencyBounded residency, hierarchy traversal, and foveated LOD.
Formats and loadersKhronos glTF splats, 3D Tiles, SPZ, Arrow, and loaders.

Limits and compatibility​

The graph renderers, GPU ordering, and storage-backed features require WebGPU. RAD hierarchy selection and generic residency can run on CPU-only data, including in an application-owned worker. Use the prepared SplatRenderer path when WebGL 2 compatibility or caller-provided source ordering is required. The package APIs remain experimental.