Over the past few weeks, we have seen attention be turned to very large scale gaussian splatting reconstructions across a variety of datasets and viewers. Joining those is JDULTRA's Contextsplat.xyz. It’s a free browser based platform built to make streaming, converting, and deploying 3DGS scenes feel seamless and scalable.

Contextsplat also tackles a more practical wish of people working with splats, converting them into formats usable in broader pipelines. Directly from the browser, you can convert Gaussian splats into OGC 3D Tiles, a format widely used for geospatial visualization and large scale 3D data delivery. For small to mid sized datasets, the conversion happens locally in your browser, with no cloud processing needed.

Once converted, you’re free to download the 3D Tiles package for further work or integration elsewhere. The tool even provides a ready to go three.js starter application, making it easier to embed your reconstructed scenes into a web project without having to start from scratch.

The platform lets you layer your 3DGS scenes with data from sources like Google 3D Tiles or OpenStreetMap, unlocking new possibilities for visualization, mapping, and context aware applications. Features like volumetric lighting play a key role in visually blending these layers into cohesive scenes, something that’s often missing from other 3DGS viewers. The platform’s OGC 3D Tiles exports use the KHR_spz_gaussian_splats_compression extension, which is the current offering from the Khronos Group.

Developers interested in experimenting with browser-based splat streaming can also use the threedtiles npm package, which supports rendering these datasets in custom projects. The geospatial industry is one that is poised to reap large scale benefits from gaussian splatting. ContextSplat is helping push the adoption and utility of 3DGS. Try it on their website!