Alexander Mai

Radiance Meshes for Volumetric Reconstruction We introduce radiance meshes, a technique for representing radiance fields with constant density tetrahedral cells produced with a Delaunay tetrahedralization. Unlike a Voronoi diagram, a Delaunay tet...

Code for the Vulkan rasterizer, WebGPU rasterizer, and training code on website.

Available now.

Website here: half-potato.gitlab.io/rm
Paper: arxiv.org/abs/2512.04076

Its not just xPBD, our radiance meshes are compatible with many different processing and rendering algorithms. To demonstrate the flexibility of radiance meshes, we extract a portion of our mesh and render it as a surface in blender:

Radiance Meshes for Volumetric Reconstruction Radiance meshes are volumetric triangle meshes, where each tetrahedron (from a Delaunay triangulation) has a constant density and linear color. These trained radiance meshes can be trivially thrown in...

We pass this order to our triangle rasterizer, which obtains the highest framerate per a primitive on MipNeRF 360 scenes, all with exact rendering. Try our online version (low quality, 50MB) here: half-potato.gitlab.io/rm/#demos

RMs generated from a Delaunay triangulation can be rasterized because the triangle mesh can be sorted front to back using the power of the circumsphere with respect to the camera origin, to obtain the correct rendering order, flickering free, within a few milliseconds.

Introducing the new fastest and most flexible view synthesis method: Radiance Meshes. RMs are volumetric triangle meshes that can be thrown into any game engine, rendered at ~200 FPS@1440p on a RTX4090, and edited using conventional tools. Here's a demo running on my desktop: