By Tamrat Belayneh (Principal Technologist, Esri), Konrad Wenzel (Director, Esri R&D Center Stuttgart GmbH), Amanda Morgan (Senior Director, Open Standards, Bentley Systems/Cesium), Adam Morris (Principal Engineer, Platform Cesium, Bentley Systems/Cesium), Azad Balabanian (Senior Product Manager Niantic Spatial, Inc)
3D Gaussian Splats are an innovative 3D representation technology emerging in the realm of reality capture. Unlike traditional solid geometries, 3D Gaussian Splats utilize radiance fields to represent the complexity of real-world environments with unprecedented photorealistic detail. This sophisticated method captures subtle features such as thin structures, semi-transparent materials, reflections, and intricate textures—previously challenging aspects for conventional 3D capture techniques and storage mechanisms popularized in photogrammetric meshes. Their unique ability to represent high-fidelity spatial content with efficient rendering has sparked broad interest across the geospatial, graphics, and standards communities.
Recognizing the transformative potential of 3D Gaussian Splats, leading organizations across the geospatial, graphics, and standards domains — including Open Geospatial Consortium (OGC), Khronos Group®, Niantic Spatial, Cesium(Bentley), and Esri— have come together to integrate 3D Geospatial Gaussian Splats into Khronos’ widely adopted glTF™ 3D asset format standard. This collaborative initiative aims to create a standardized, interoperable framework for encoding and sharing 3D Gaussian Splats, ensuring broad interoperability and easy implementation across platforms and applications.
SPZ and glTF: Efficiency, Simplicity, and Flexibility
Central to this initiative is the adoption of the SPZ format, an open-source file format provided by Niantic Spatial under the MIT License, which compresses 3D Gaussian splats by up to 90% compared to PLY while preserving visual fidelity and performance. The simplicity and efficiency of SPZ make it ideally suited for widespread adoption, as it strikes a balance between computational performance and high-quality visualization.
Combining SPZ’s strengths with glTF’s flexible structure will enable an interoperable and straightforward implementation that allows for future extensibility via the potential inclusion of additional data fields to accommodate diverse user requirements and custom workflows.
New Gaussian Splatting glTF Extensions in Development
The Khronos 3D Formats Working Group is now developing two new extensions, aimed at standardizing the delivery of Gaussian Splats within glTF assets:
- KHR_gaussian_splatting, which defines the structure for storing 3D Gaussian splats in glTF, treating them as point primitives with attributes such as position, rotation, scale, transparency, and spherical harmonics (supporting both diffuse and specular components). This structure also allows graceful fallback to sparse point cloud rendering.
- KHR_gaussian_splatting_compression_spz, which enables efficient storage and streaming using the SPZ format. SPZ blobs are stored as buffers within glTF primitives and can be decompressed into attributes or passed directly into rendering pipelines. It supports flexible encoding of spherical harmonics—ranging from none to three degrees—based on content needs.
These extensions are forward-looking and intended to serve as a foundational pathway for long-term support of 3D Gaussian splatting in glTF. They aim to provide an extensible, performant base layer for high-fidelity spatial rendering, with room to grow into more complex capabilities as the technology matures.
To validate the approach, the collaborative group conducted comprehensive evaluations across a variety of geospatial datasets. A recurring challenge emerged in scenarios involving long, linear features common in geospatial contexts—such as antennas, fences, power lines, and rail tracks—which exhibited visual artifacts that were unsuitable for accurate visualization and analysis. Capturing these elongated splats without significantly increasing payload size proved difficult.
To address this, the group introduced a minimal yet critical enhancement to rotational accuracy in the SPZ format. which was recently released as Version 2.0.0 of the SPZ library. With this change, rotations in SPZ are now encoded using the smallest three components of a normalized quaternion, each stored as a 10-bit signed integer, while the largest component is derived and its index stored using 2 bits—optimizing for accuracy. In contrast, previous versions of SPZ used the fixed (x, y, z) components of the quaternion, also omitting the derived w component, but with less precise encoding.
Follow-up evaluations demonstrated substantial improvements in data quality, reinforcing the practicality and adaptability of the solution for real-world geospatial applications.
Get Involved
As this work evolves, we actively welcome feedback and contributions from the broader community to help shape the newly proposed glTF extensions. Whether you’re a researcher, developer, standards contributor, or 3D practitioner, your insights are essential to building a robust specification that addresses real-world needs. All are welcome to become a Khronos Group Member and take a direct role in the development of extensions by participating in the Khronos 3D Formats Working Group. You can also contribute by engaging in discussions on OGC standards, and open collaboration on GitHub:
About Khronos
The Khronos Group is an open, non-profit, member-driven consortium of over 150 industry-leading companies creating royalty-free, interoperability standards for 3D graphics, augmented and virtual reality, parallel computation, vision processing and machine learning. Khronos activities include 3D Commerce™, ANARI™, glTF™, NNEF™, OpenCL™, OpenGL®, OpenGL® ES, OpenVG™, OpenVX™, OpenXR™, SPIR-V™, SYCL™, Vulkan®, and WebGL™. Khronos members drive the development and evolution of Khronos specifications and are able to accelerate the delivery of cutting-edge platforms and applications through early access to specification drafts and conformance tests.
About the Open Geospatial Consortium (OGC)
The Open Geospatial Consortium (OGC) is a global, cross-sector community of more than 500 organizations working to advance open standards and innovation in geospatial technology. For over 30 years, OGC has united industry, government, academia, and research to solve complex challenges through interoperability, collaboration, and real-world testing. From climate resilience and disaster response to smart cities, digital twins, autonomous systems, and advanced 3D visualization, OGC standards and innovation initiatives enable the seamless sharing and use of geospatial information across platforms and domains. By connecting diverse expertise and technologies, OGC helps organizations make better decisions, improve quality of life, and create a more resilient, sustainable future. Learn more at www.ogc.org.
About Esri
Esri, an influential member of the OGC and a global leader in geospatial solutions for more than five decades. With extensive expertise in addressing complex geospatial challenges, Esri continues to drive innovation aimed at facilitating seamless access to massive geospatial datasets across diverse platforms—including web browsers, mobile devices, and desktop applications. This collaborative advancement further underscores Esri’s enduring commitment to enhancing geospatial interoperability and accessibility, setting new benchmarks in geospatial data representation and usage.
About Bentley Systems (Cesium)
Cesium is the platform for 3D geospatial. We created 3D Tiles, the OGC community standard for streaming massive 3D geospatial data. Creators use Cesium to build with real-world data at scale, across industries, including AEC, aerospace, defense, environment, telecommunications, and more. Bentley Systems acquired Cesium in September 2024. Founded in 1984 by engineers for engineers, Bentley is the partner of choice for engineering firms and owner-operators worldwide, with software that spans engineering disciplines, industry sectors, and all phases of the infrastructure lifecycle.
About Niantic Spatial
Niantic Spatial is a pioneer in geospatial AI, building technology that enables both people and machines to perceive and understand physical spaces in entirely new ways. Our core technology is built on a third-generation digital map, and the Visual Positioning System (VPS) delivers centimeter-level precision in real-world localization. We are developing a Large Geospatial Model (LGM) to deliver a semantically rich, spatially grounded understanding of real-world locations.
Registered Trademarks:
Khronos® and Vulkan® are registered trademarks, and ANARI™, WebGL™, glTF™, KTX™, NNEF™, OpenVX™, SPIR™, SPIR-V™, SYCL™, OpenVG™, Karamos™ and 3D Commerce™ are trademarks of The Khronos Group Inc. OpenXR™ is a trademark owned by The Khronos Group Inc. and is registered as a trademark in China, the European Union, Japan and the United Kingdom. OpenCL™ is a trademark of Apple Inc. and OpenGL® is a registered trademark and the OpenGL ES™ and OpenGL SC™ logos are trademarks of Hewlett Packard Enterprise used under license by Khronos. All other product names, trademarks, and/or company names are used solely for identification and belong to their respective owners.