Registration is open for the 2026 Cesium Developer Conference! Save your spot.

Cesium for Omniverse v0.28.0

Cesium for Unity Samples v1.23.2

CesiumJS 1.142

Cesium ion

人工衛星が取得するデータ量は急速に増大している。将来的には、地球観測衛星だけでも1日あたりペタバイト級のデータが生成されるとも予測されている。こうした膨大なデータをすべて地上に送信してから処理する従来の方式では、通信帯域や地上局数の制約から限界に近づきつつある。 

こうした課題に対し、NTTソフトウェアイノベーションセンタでは「宇宙データセンタ」と呼ばれるコンピューティングアーキテクチャの研究開発が進められている。この取り組みでは、人工衛星上でAI推論などのデータ処理を実行し、重要な情報のみを地上に送信することで、通信効率の向上と迅速なデータ活用を目指している。 

このアーキテクチャの設計や検証を支える可視化の試みとして、Webベースの3D地球可視化エンジンであるCesiumJSが活用されている。

宇宙でAIが動く新しいデータ処理モデル 

その中核の一つとなるのが、衛星軌道および衛星間通信リンクを統合的にシミュレーションできる基盤である。 

従来の地球観測ワークフローは、次のような流れで構成されていた。 

これに対し、NTTが検討している新しいモデルでは、 

重要な情報のみを地上へ送信 
という処理アーキテクチャが採用されている。  

NTTの技術展示イベント「NTT R&Dフォーラム2025」で公開されたデモでは、人工物検知アルゴリズムを用いて重要領域のみを抽出することで、送信データ量を約96%削減できる可能性が示された。 

ハイパースペクトルデータ（疑似RGB画像）: 衛星上で取得された観測データ（ハイパースペクトルデータ）の可視化例。（提供：NTT）

AI処理による重要領域抽出結果（データ削減後）: 同データに対してAI処理を適用し、重要または異常と判断された領域のみを抽出した例。（提供：NTT）

このアプローチは単なる通信効率の改善にとどまらない。 
限られた通信帯域の中で「どの情報を優先して地上へ送るか」を選別できるため、 

インフラの異常検知 
といった迅速な判断が求められる分野での応用が期待されている。  

宇宙データセンタのアーキテクチャでは、単一の衛星ではなく、複数の要素が連携する分散システムが想定されている。 
例えば次のような要素が同時に関係する。 

時間とともに変化するネットワーク構造  

宇宙データセンタの概念を可視化したもので、衛星上でのデータ処理の様子を示している。（提供：NTT） 

Visualization of a space data center concept, showing satellite-based processing in orbit. Courtesy NTT. 
宇宙データセンタの概念を可視化したもので、衛星上でのデータ処理の様子を示している。（提供：NTT）

NTT: space data center concept

NTTでは、宇宙データセンタを構成する衛星群の軌道や通信リンク、軌道上でのAI処理を統合的にシミュレーションできるライブラリも開発している。しかし、衛星数やシミュレーション期間が増えるにつれて、可視化処理の負荷が増大し、研究環境の制約になるケースも見られた。 

この課題に対し、NTTでは大規模な軌道シミュレーションのための可視化環境の構築・検証にCesiumJSを用いた。 

CesiumJSを用いた宇宙システムの可視化 

NTTが可視化の選択肢を比較検討する中で、CesiumJSのいくつかの特徴が有効であることが分かった。 

CZMLによる時間変化を伴うオブジェクト表現 
衛星軌道や通信リンクを自然に表現できる  

Webブラウザ上での高性能レンダリング 
大規模シミュレーションでも操作性を維持できる  

地理空間データとの統合 
地形や衛星画像と組み合わせて地球上の文脈で理解できる  

CesiumJSを用いることで、衛星の軌道、通信リンク、観測対象地点などを3D地球上で直感的に把握できる可視化環境が実現された。 

分散ネットワークアーキテクチャにおける衛星間通信リンクの可視化。（提供：NTT） 

Inter-satellite communication links visualized in a distributed network architecture. Courtesy NTT. 
分散ネットワークアーキテクチャにおける衛星間通信リンクの可視化。（提供：NTT）

NTT: inter-satellite communication links

また、Cesium ionの地理空間データを組み合わせることで、宇宙空間の挙動と地球上の観測対象を同時に把握することが可能になっている。

可視化・分析レイヤとしてのCesium 

本プロジェクトにおいて、Cesiumは可視化および分析のレイヤーとして機能する。 

オンボード処理パイプラインの構築や、前処理・AI推論フローの設計を担っているのは、NTTが開発した low-code / no-code のパイプライン設計ツールである。 

CesiumJS は、そうした処理構成や実行結果を3次元空間上で可視化・分析するためのレイヤとして機能している。衛星軌道、通信リンク、観測地点、センサー由来データやAI推論結果を地形やインフラと統合して表示することで、研究者やエンジニアはシステム全体の挙動を直感的に把握できる。 

Cesiumはパイプラインを「設計する」ためのインタフェースではない。一方で、複雑な宇宙分散システムがどのように振る舞っているかを“観測可能”にし、仮説検証や異常の発見、関係者間での認識共有を支援する。 

このようにCesiumは、NTTの設計ツール群と連携しながら、研究開発を加速する可視化・分析基盤として重要な役割を果たしている。 

研究開発を加速する可視化基盤 

可視化部分をCZMLベースの構成に整理したことで、ソフトウェア構成はシンプルになり、開発効率も向上した。 
その結果、研究チームは開発開始から約2か月でプロトタイプをNTT R&Dフォーラムに展示することができた。 

可視化の実装やレンダリング最適化にかかる負担が軽減されたことで、研究者はより多くの時間を、軌道計算や通信リンク成立判定、分散処理アーキテクチャ設計などの宇宙データセンタの実現に向けた中核機能の研究開発に集中できるようになった。 

現在のプロトタイプでは、主にCZMLを用いた衛星軌道や通信リンクの可視化が中心となっている。将来的に衛星コンステレーション規模が拡大し、より大規模なデータ可視化処理が必要になった場合には、3D Tilesなどの階層的ストリーミング技術の活用も検討されている。 

宇宙で処理し、地上で理解する。その橋渡し役として、空間可視化技術の重要性は今後ますます高まっていくだろう。 

Cesiumでは、NTTのように新しい宇宙コンピューティングアーキテクチャを探求する組織が、複雑なシステムを理解し共有するための基盤としてCesiumを活用していることを嬉しく思います。 

他の航空宇宙企業のCesium活用例についてもぜひご覧ください。 

NTT Explores a Space Data Center Architecture with Cesium for On-Orbit AI and Satellite Network Visualization

The volume of data collected by satellites is increasing rapidly. In the future, even Earth observation satellites alone are expected to generate data at the petabyte-per-day scale. Under the conventional approach of transmitting all of this data to the ground before processing, limitations in communication bandwidth and the number of available ground stations are becoming increasingly significant. 

To address this challenge, NTT Software Innovation Center is conducting research and development on a computing architecture referred to as the “Space Data Center.” In this approach, data processing such as AI inference is executed onboard satellites, and only important information is transmitted to the ground, with the goal of improving communication efficiency and enabling faster use of data. 

As part of the effort to support the design and validation of this architecture, NTT is using CesiumJS, a web-based 3D globe visualization engine. 

A new data processing model in which AI operates in space 

One of the core elements is a platform capable of integrated simulation of satellite orbits and inter-satellite communication links. 

A conventional Earth observation workflow has typically consisted of the following steps: 

A satellite observes Earth’s surface.  

Analysis and processing are performed on Earth.  

By contrast, in the new model being explored by NTT, the processing architecture is: 

Only important information is transmitted to the ground.  

In a demonstration presented at NTT’s technology showcase event, NTT R&D Forum 2025, an object-detection algorithm was used to extract only important regions, showing the possibility of reducing transmitted data volume by approximately 96%. 

Hyperspectral data (pseudo-RGB visualization): Example of observation data (hyperspectral data) acquired onboard a satellite, visualized as a pseudo-RGB image. Courtesy NTT. 

AI-processed result highlighting regions of interest (after data reduction): Example of applying AI processing to the same dataset, extracting only regions identified as significant or anomalous. Courtesy NTT. 

This approach is not limited to improving communication efficiency alone. 

Because it makes it possible to select which information should be prioritized for transmission to the ground within limited communication bandwidth, it is expected to be applicable in fields requiring rapid decision-making, such as: 

A distributed system expanding into space 

The Space Data Center architecture assumes not a single satellite but a distributed system in which multiple elements work together. 

For example, the following elements are involved simultaneously: 

network structures that evolve over time.  

Visualization of a space data center concept, showing satellite-based processing in orbit. Courtesy NTT. 

NTT has also developed a library capable of integrated simulation of the satellite orbits, communication links, and onboard AI processing involved in the Space Data Center. However, as the number of satellites and the simulation duration increased, the computational load of visualization also increased, becoming a constraint in the research environment in some cases. 

To address this issue, NTT conducted development and validation of a visualization environment using CesiumJS for large-scale orbital simulations. 

Visualizing space systems with CesiumJS 

As NTT evaluated different visualization options, several characteristics of CesiumJS proved effective: 

Object representation with temporal changes using CZML: 
Orbital paths and communication links can be represented naturally.  

High-performance rendering in a web browser: 
Operability can be maintained even in large-scale simulations.  

Integration with geospatial data: 
Space systems can be understood in the context of Earth by combining them with terrain and satellite imagery.  

By using CesiumJS, NTT created a visualization environment in which satellite orbits, communication links, and observation target locations can be intuitively understood on a 3D globe. 

Inter-satellite communication links visualized in a distributed network architecture. Courtesy NTT. 

In addition, by combining it with geospatial data from Cesium ion, it becomes possible to understand space-based behavior and Earth-based observation targets at the same time. 

In this project, Cesium plays a central role as a visualization and analysis layer. 

NTT’s low-code/no-code pipeline design tool provides the primary interface for configuring onboard processing workflows, including preprocessing steps and AI inference pipelines. 

CesiumJS serves as the visualization and analysis layer that complements this design environment. By visualizing system configurations, sensor data, and processing outputs in a unified 3D context, Cesium enables researchers and engineers to intuitively 

inspect, validate, and reason about complex onboard behaviors more effectively. 

Rather than acting as a pipeline authoring interface, Cesium focuses on making system behavior observable.   

This visualization-driven analysis shortens iteration cycles by helping researchers identify anomalies, evaluate assumptions, and communicate findings across interdisciplinary teams. In this way, Cesium provides a foundational layer that accelerates R&D while integrating seamlessly with NTT’s broader toolchain. 

A visualization foundation that accelerates R&D 

By reorganizing the visualization component into a CZML-based structure, the software configuration became simpler and development efficiency improved. 

As a result, the research team was able to demonstrate a prototype at NTT R&D Forum, in approximately two months from the start of development. 

By reducing the burden associated with visualization implementation and rendering optimization, researchers were able to spend more time focusing on the research and development of core functions required to realize the Space Data Center, such as orbital calculations, communication link validation, and distributed processing architecture design. 

In the current prototype, visualization mainly focuses on satellite orbits and communication links represented using CZML. In the future, as satellite constellation scale increases and larger-scale visualization processing becomes necessary, the use of hierarchical streaming technologies such as 3D Tiles is also being considered. 

Process in space, and understand on Earth. As a bridge between the two, the importance of geospatial visualization technology is expected to grow even further. 

At Cesium, we are pleased to see organizations like NTT—exploring new space computing architectures—use Cesium as a foundation for understanding and sharing complex systems. 

Please also explore how other aerospace companies are using Cesium. 

 Cesiumで可視化する軌道上AIと衛星ネットワーク ――  NTTが探る「宇宙データセンタ」アーキテクチャ

In May, we attended GEOINT Symposium in Aurora, Colorado, USA, and OGC Builder Days in London, UK.  

We’re ramping up to this year’s Cesium Developer Conference—next week!—June 8-10 in Philadelphia. Have you registered to join us? Tracks include Aerospace, Defense, and Simulation; Cesium Core: Product Roadmaps and Advanced Technologies; Digital Twins and AEC Innovation; and Next-Gen Data and 3D Workflows. Learn about 3D Tiles 2.0, vector data, voxels, design review, sensor fusion, emergency management, synthetic training environments, Mars, and more. We look forward to connecting with you at the pre-conference workshops, in the ionosphere, during talks, 1:1, and at receptions, to shape the future of open geospatial together.

Join us for the 2026 Cesium Developer Conference next week.

We made improvements across Cesium this month. To get notifications about our monthly releases, you can subscribe to the Cesium release roundup thread on our community forum.

CesiumJS 1.142 is now available. Highlights of the release include:

Rendering support for vector tiles from 3D Tiles generated by the Cesium ion vector tiles technology preview. Includes support for 3D points, lines, and polygons, with property-based styling and picking. Includes 3DTILES_content_gltf_vector extension for 3D Tiles, and KHR_mesh_primitive_restart and EXT_mesh_primitive extensions for glTF. For more information, see 3d-tiles#825. #13478 

Sandcastle now ships with Cesium Copilot, an in-app AI coding assistant that writes, edits, and debugs CesiumJS code in the same panel as the live preview. Users bring their own keys for Google Gemini, Anthropic Claude, or Google Vertex AI; the assistant edits the JavaScript and HTML buffers through a structured apply_diff tool, streams thinking and tool calls live, and can self-fix runtime errors without a manual round-trip. #13188 

Added GeoJsonPrimitive for loading GeoJSON directly into BufferPrimitiveCollections, bypassing the entity/DataSource layer for significantly improved performance with large datasets. #13505 

Added MVTDataProvider for loading Mapbox Vector Tiles (MVT) directly into CesiumJS as 3D Tiles. Supports per-feature styling via Cesium3DTileStyle, feature picking with metadata (getProperty), and automatic property table encoding via EXT_structural_metadata. #13404 

- BufferPrimitiveCollection improvements: 
    - Added `boundingVolume` constructor parameter to `BufferPointCollection`, `BufferPolylineCollection`, and `BufferPolygonCollection`. For larger animated collections, providing a precomputed bounding volume can eliminate the performance cost of automatically updating the bounding volume frequently. [#13477](https://github.com/CesiumGS/cesium/pull/13477) 
    - `boundingVolume` property is now defined in world space, not local/model space. [#13477](https://github.com/CesiumGS/cesium/pull/13477)  
    - Added `.blendOption` constructor parameter to BufferPrimitiveCollection. 
    - Added `BufferPrimitiveMaterial#color.alpha`. [#13384](https://github.com/CesiumGS/cesium/pull/13384)
    - Added `BufferPointMaterial#outlineColor.alpha`. [#13384](https://github.com/CesiumGS/cesium/pull/13384) 
    - Fixed a bug causing `BufferPointCollection` to not update after changes to point positions. [#13465](https://github.com/CesiumGS/cesium/pull/13465) 

Added support for multiple key modifiers in ScreenSpaceEventHandler.setInputAction. #13307 

Added EdgeDisplayMode enum and edgeDisplayMode property to Model and Cesium3DTileset for controlling how edges from the EXT_mesh_primitive_edge_visibility glTF extension are rendered. Supports three modes: SURFACES_ONLY, SURFACES_AND_EDGES, and EDGES_ONLY (CAD-style wireframe rendering). #13192 

3D vector tiles in CesiumJS visualizing building footprints categorized by construction year in New York City, viewed in Cesium Sandcastle.

Cesium for Unreal v2.27.0 is now available. Highlights of this release include:

Added CesiumVectorTilesRasterOverlay supporting vector data loaded from 3D Tiles tilesets. 

Added support for displaying points in CesiumGeoJsonDocumentRasterOverlay and CesiumVectorTilesRasterOverlay. 

Added support for rendering edges from the EXT_mesh_primitive_edge_visibility extension. 

Added missing includes that introduced compilation failures when building the plugin from source against UE 5.7's bundled clang 20.1.8 toolchain (introduced in v2.25.0 by #1685; binary plugin users were unaffected). 

Fixed the case where glTF line primitives caused excessive log spam due to Chaos attempting to generate physics bodies for small non-triangle meshes. 

Fixed an incorrect transform that reversed the appearance of data in box voxel tilesets. 

Fixed a crash in UCesiumGaussianSplatSubsystem::Tick when GetPrimaryWorld() returns nullptr (e.g., between PIE shutdown and Standalone-Game launch, or on nDisplay secondary nodes during world bind-up). Restores the validity check ordering that shipped in v2.24.1 and was inadvertently inverted in v2.26.0 (#1841). 

In addition to the above, this release updates cesium-native from v0.60.0 to v0.61.0. See the changelog for a complete list of changes in cesium-native. 

Check the Cesium for Unreal release and the Cesium for Unreal Samples release for the full list of updates.

Download the Cesium for Unreal plugin from Epic Marketplace

New York City building footprints tiled as 3D Tiles through Cesium ion, streamed in Cesium for Unreal using the new vector tiles raster overlay. 

Cesium for Unity v1.23.3 is now available. Highlights of this release include:

The native Android library (libCesiumForUnityNative.so) is now built with 16 KB ELF segment alignment, enabling compatibility with Android 15+ devices that use 16 KB memory pages. 

In addition to the above, this release updates cesium-native from v0.60.0 to v0.61.0. See the changelog for a complete list of changes in cesium-native.  

Check the Cesium for Unity changelog for the full list of updates, and follow our Cesium for Unity tutorials to get started. 

**Cesium ion SaaS**
- Cesium ion no longer requires mesh generation for reconstruction workflows, enabling point cloud-only and Gaussian splat-only jobs to run through the reconstruction pipeline. 
- Coming this week: expanded support for design model formats, making it easier than ever to bring engineering, architecture, and infrastructure data into 3D Tiles to stream anywhere. 
- Model tiler will be retired from Cesium ion on September 1, 2026. It will no longer be available through the Cesium ion user interface or API. Other tilers produce better results for the file formats supported and output 3D Tiles 1.1. This tiler outputs 3D Tiles 1.0. 

**Vector tiler** 
- Coming this week: support for tiling large .geojson files via the new vector tiler technology preview.

**Voxel tiler**
- Coming this week: support for tiling large NetCDF files via the new voxel tiler technology preview. 

The Cross Product built TCP Studio to visualize and analyze large LiDAR datasets for linear infrastructure, using CesiumJS and Cesium ion to stream 3D Tiles at scale. 

Neuroscientist Lu Zhang created EarthRunner, an immersive treadmill experience that allows users to run in locations all over the world during a home workout, using Cesium for Unreal to stream Google Photorealistic 3D Tiles of real-world locations. 

In April, we announced the 2026 Cesium Developer Conference tracks, published the full agenda, and highlighted the pre-conference workshops; excitedly tracked Artemis II; made 3D Gaussian splats available in the Cesium and 3D Tiles ecosystems; and shared what we’re showcasing at GEOINT Symposium next week. Will we see you there? Join us in booth 1303. 

CesiumJS 1.141 is now available. Highlights of the release include:

Added support for properties (EXT_structural_metadata) in vector tilesets. #13426 

Bumped minimum required Node version to 22.0.0 .

BufferPrimitiveCollection properties modelMatrix, boundingVolume, and boundingVolumeWC are now readonly. They may be modified, but not reassigned. #13448 

Fixed lighting affecting EquirectangularPanorama. #13369 

Fixed incorrect matrix multiplication for non worldspace instance transforms in pickModel. #13433 

A 3D Gaussian splat tileset of the Microsoft campus in Redmond, Washington, USA, in CesiumJS. (Explore the Sandcastle.)

Cesium for Unreal v2.26.0 is now available. Highlights of this release include:

Added ACesium3DTileset::UnloadEditorTilesInPlayMode, which toggles whether in-editor tileset instances should unload their tiles before Play-in-Editor begins. This can prevent the Editor from duplicating resources for the same tileset and potentially consuming too much memory. 

Added a "Clear Request Cache" Blueprint-callable function to UCesiumRuntimeSettings and a button to Project Settings -> Plugins -> Cesium that clears all entries from the SQLite request cache database. 

Fixed a bug where Gaussian splats would continually accumulate in the scene, eventually causing the application to crash. 

In addition to the above, this release updates cesium-native from v0.59.0 to v0.60.0. See the changelog for a complete list of changes in cesium-native. 

The Microsoft campus in Redmond, Washington, USA, visualized with 3D Gaussian splat tilesets in Cesium for Unreal. 

Cesium for Unity v1.23.2 is now available. 

This release updates cesium-native from v0.59.0 to v0.60.0 and brings bug fixes such as:

Fixed a bug in CesiumRasterOverlays::GeoJsonDocumentRasterOverlay where styles set on geometry objects within a FeatureCollection were not applied.   

See the changelog for a complete list of changes in cesium-native. 

**Cesium ion SaaS**
- Cesium ion now enables organizations with paid accounts to require the use of SSO when their users log into ion. To learn more, contact support. 
- Cesium ion now generates levels of detail (LODs) for Gaussian splat assets generated through the reconstruction pipeline. 

**Cesium ion Self-Hosted** 
Cesium ion Self-Hosted 1.10.0 is now available. 

- Added support for ingestion and streaming of 3TZ assets. 
- Added UI enhancements to the Cesium ion dashboard. 
- Design tiler: Improved rendering of IFC elements by giving sub-assembly elements the bounding box of the full assembly. 
- Design tiler: Added more graceful handling of IFC lines that don’t meet the IFC specification. 
- Design tiler: Added support for parsing eastings, northings, and orthogonal height from IFC files. 
- Gaussian splats tiler: Added logic to filter out NaN and infinite values when parsing PLY files. 
- Design tiler: Emits double-precision floats (instead of single-precision floats) for IFC attributes and properties. 
- Design tiler: Fixed generation of IFC swept circular geometry for better shading and performance. 
- Design tiler: Fixed issues processing B-splines in IFC files. 
- Design tiler: Fixed issues processing some geometries with swept circles. 
- Reality tiler: Fixed an issue where duplicate vertices in input models could lead to inefficient textures and larger tilesets than necessary. 
- Reality tiler: Fixed a regression where duplicate vertices in the input would be written to tiles, rather than being filtered out. 
- Reality tiler: Significantly reduced memory usage for models with spatially incoherent textures. 

**Asset Server**
- Updated the asset server to the latest version. 

**Design Tiler**
- Added `IfcSpace` relations to element properties in IFC metadata. 

**Reality Tiler** 
- Fixed a race condition that could lead to subtree files with content/availability bits incorrectly set to zero. 

Bedrock.engineer leverages voxel support in draft 3D Tiles 2.0 and CesiumJS to stream large-scale subsurface geoscience data directly in the browser for city planning, infrastructure, and environmental decision-making.

XReality Factory (XRF) uses CesiumJS and Cesium for Unity for a single shared operational view for better, faster decisions across real operations and training scenarios, throughout defense, security, and emergency management. 

The Purdue University Envision Center has integrated Cesium for Unity into CollabXR, bringing geospatial support to their XR development platform that allows students and researchers to explore complex datasets together. 

Geotour brings together information about Crete’s history, landscapes, and biodiversity as well as photos, panoramas, and video tours of the island in a single interactive map built using CesiumJS. 

3D Gaussian splats are now available in the Cesium and 3D Tiles ecosystems, including CesiumJS, Cesium for Unreal, and Cesium ion. This integration enables developers to stream and visualize high-fidelity, photorealistic 3D content that captures the real world with unprecedented detail. By leveraging 3D Tiles as a spatial index and glTF as the payload, massive 3D Gaussian splat datasets can now be streamed with level of detail (LOD), ensuring high performance from city-scale environments down to sub-centimeter details. 

3D Gaussian splatting is an efficient radiance field reconstruction method that converts 2D images into a sparse cloud of volumetric 3D Gaussians. Unlike traditional photogrammetry, which often struggles to represent thin or complex structures, 3D Gaussian splats excel at preserving visual fidelity for cables, power lines, vegetation, and reflective surfaces by creating a point cloud of soft, rounded shapes of the subject, with each point defined by properties including position, orientation, color, and opacity. This technique uses spherical harmonics to support view-dependent lighting, allowing for realistic specular reflections and anisotropy that traditional reality meshes cannot easily replicate. 

3D Gaussian splats are now available in the Cesium and 3D Tiles ecosystems, including CesiumJS, Cesium for Unreal, and Cesium ion.

3D Gaussian Splats with Hierarchical Level of Detail Using 3D Tiles

3D Gaussian splatting provides significant advantages for industries requiring high-fidelity reality capture and photorealistic digital twins, including infrastructure monitoring, telecom tower maintenance, and electrical utility and substation inspections—plus the built environment, such as in real estate, where traditional mesh and point cloud methods often have limits. Beyond industrial applications, we see 3D Gaussian splats being adopted in media and entertainment to capture challenging visual details like reflections, semitransparent materials, and fire. 3D Gaussian splats are expected to play a key role in long-term asset monitoring and situational awareness, enabling users to track changes in the natural and built environment over the entire life cycle. 

3D Gaussian splats enable high-fidelity accurate visualizations of thin-and-long assets like cranes, building materials, power stations, and cell towers. This crane is depicted with Cesium for Unreal.

3D Gaussian splats enable high-fidelity accurate visualizations of thin-and-long assets like cranes, building materials, power stations, and cell towers. This scene is depicted in CesiumJS.

Standardizing 3D Gaussian splats in glTF with SPZ compression 

To ensure 3D Gaussian splats are interoperable across platforms, we have collaborated with Khronos, Open Geospatial Consortium (OGC), Esri, and Niantic Spatial to integrate this new graphics primitive into the glTF 3D asset standard. This effort centers on two new extensions: KHR_gaussian_splatting and KHR_gaussian_splatting_compression_spz. The base glTF extension defines how splats are stored as point primitives, which allows for a graceful fallback to sparse point clouds in renderers that do not yet support rendering 3D Gaussian splats. 

"The geospatial community has been the catalyst for bringing Gaussian splats into glTF as an open, interoperable primitive, and the close collaboration between Cesium, Esri, and Niantic Spatial at OGC and Khronos is a model for how new rendering techniques should enter the industry: quickly, openly, and with broad cross-community backing from day one. Ratification at Khronos elevates Niantic's SPZ compression from an open-source project into a royalty-free open standard, providing the IP certainty that enterprises and platforms need to adopt it at scale." 

—Neil Trevett, President, Khronos Group 

A major challenge for 3D Gaussian splats has been their massive file size, which can hinder efficient streaming. To solve this, we adopted the SPZ compression format, an open source solution provided by Niantic Spatial. SPZ uses advanced quantization techniques and gzip to compress splat data by up to 90% compared to standard PLY files while maintaining visual fidelity. This format is particularly effective at handling spherical harmonic data for specular lighting. Furthermore, the recent SPZ 2.0.0 release introduced improved rotational accuracy for linear geospatial features—such as antennas and power lines—by encoding rotations as normalized quaternions. 

3D Gaussian splats will be added to 3D Tiles as part of the proposed 3D Tiles 2.0 OGC community standard. 

"The Open Geospatial Consortium (OGC) endorsed 3D Tiles as a Community Standard because it offered a compelling ability to distribute 3D content to users in a performant manner that also adhered to best practices in handling geospatial information and accuracy. This same approach to provide Gaussian splats via glTF and 3D Tiles ensures that the information portrayed by those Gaussian splats is in the right place and provides meaningful context for geospatial analysis."  

— Scott Simmons, Chief Standards Officer, OGC

Creating 3D Gaussian splats using Cesium ion 

With the integration of iTwin Capture technology into Cesium ion, the platform provides an end-to-end pipeline for reality modeling, including the creation of 3D Gaussian splats with LOD. Users can upload source photos directly to Cesium ion, which will automatically reconstruct a 3D Tileset as a mesh, point cloud, or 3D Gaussian splats. 

To accommodate different development needs, these workflows are accessible through both the Cesium ion web interface and via REST APIs, enabling everything from manual one-off uploads to fully automated, high-scale production pipelines. Once tiled, these splats are georeferenced, allowing them to be instantly combined with global datasets like Cesium World Terrain or Google Photorealistic 3D Tiles for full geospatial context. 

3D Gaussian splat tileset of Microsoft Redmond Campus, as viewed in CesiumJS. Captured in partnership with Bentley Systems. This tileset uses 20,169 photos (427.7 gigapixels, DJI FC6310 camera) and covers an area of about 3.7 sq km, with mean ground sampling distance 3 cm. The reconstruction is 110 million splats.

Streaming to CesiumJS, Cesium for Unreal, and more 

CesiumJS now supports 3D Gaussian splat tilesets. The implementation uses specialized shaders for volumetric rendering and high-performance sorting using WebAssembly for increased performance. Developers using CesiumJS can build high-fidelity digital twin applications leveraging 3D Gaussian splat tilesets in conjunction with previously supported data types. 

"The barriers to working with Gaussian splatting continue to fall. With a glTF extension and LOD support now part of 3D Tiles, developers can increasingly work with splats within familiar tools and workflows. Gaussian splats are quickly becoming a professional, scalable medium for delivering lifelike 3D on the web."  

— Michael Rubloff, Founder and Managing Editor, RadianceFields.com 

For developers building immersive experiences with game engines, Cesium for Unreal now supports 3D Gaussian splat tilesets, as well. This enables the integration of massive 3D Gaussian splat dataset rendering into the Unreal Engine ecosystem with the ability to stream 3D Gaussian splats with LOD directly from Cesium ion. In addition to Cesium for Unreal, we plan to add support to Cesium for Unity and other runtime engines based on community feedback. By standardizing splat delivery through 3D Tiles and glTF extensions, we’re ensuring that these assets remain interoperable across platforms as the field of neural rendering evolves. 

Getting started with 3D Gaussian splats 

Upload your own photos to Cesium ion by signing up for a free community account. You can try out live examples in the Cesium Sandcastle or start building your own reality models today. 

Introducing 3D Gaussian Splats with Hierarchical Level of Detail Using 3D Tiles

In complex, high-pressure environments, how information is presented can be as important as the information itself. XReality Factory (XRF) uses CesiumJS and Cesium for Unity for a single shared operational view for better, faster decisions across real operations and training scenarios, throughout defense, security, and emergency management.

XRF's tools consolidate data sources and connect agencies via a single shared operational view with 3D geospatial context for better decisions. Courtesy XRF.

XReality Factory

The volcanic eruptions on the Spanish island of La Palma in 2021 were the tipping point for XRF’s founding team, with expertise in extended reality and interactive technologies, to build tools to consolidate data sources and connect agencies via a single shared operational view with 3D geospatial context for better decisions.

“What we were developing in gaming is useful in emergencies, useful to our heroes,” said Raul Espada, XRF’s CRO.

XRF supports users in critical infrastructure, defense, emergency management, and security, working with the Spanish Armed Forces, Royal Netherlands Armed Forces, public authorities, and industry. In NATO exercises with armed forces from multiple nations on different systems, XRF tools served as the common operating picture (COP), enabling coordination across distributed teams, improving situational awareness, and accelerating decision-making with meaningful information, instead of a barrage of siloed data. The company is part of NATO’s EU Defence Innovation Scheme (EUDIS) and Defence Innovation Accelerator for the North Atlantic (DIANA) programs.

In NATO exercises, XRF tools served as the COP, enabling coordination across distributed teams. Courtesy XRF.

At the core of XRF’s dual-use command and control (C2) system is Compass, a platform for fusing data from several sources into a trusted 3D COP. With Cesium World Terrain or Cesium World Bathymetry via Cesium ion as Compass’s base, users add their own 3D models, drone imagery, and other information layers specific to their missions and organizational requirements for rapid, efficient delivery, streaming as 3D Tiles into CesiumJS for the web and Cesium for Unity for XR.

Around Compass, XRF developed complementary tools:

Apolo, a mobile app for units deployed on the ground, built with CesiumJS;

Colossus, a large multitouch command table for fixed control rooms, using CesiumJS;

Sandbox, an XR multiuser command center for planning, training, and operations, built on Cesium for Unity and device agnostic across Microsoft HoloLens, Meta Quest, and Apple Vision Pro; and

Hermes, XRF’s AI system that interprets scenarios to support decisions with natural language.

With real-world data available on a variety of devices, users can more quickly understand terrain, elevation, line of sight, and spatial relationships in context.

“Our goal has always been to integrate the solution the final user needs in their unifying system, so most of our effort goes into interoperability,” said Antonio Sanchez, XRF’s primary developer.

3D Tiles, an OGC community standard built on glTF and other 3D data types, enables XRF’s users to efficiently stream massive datasets, at any scale, together. The view their users need is the one that will load with the most detail, supporting immersive visualization and analysis on both high-end workstations and bandwidth-constrained devices.

XRF uses Cesium ion SaaS for testing, demos, and situations with reliable internet. The team and its users employ Cesium ion Self-Hosted for strict internal security requirements and off-grid environments.

“The level of performance—rendering continental-scale terrain down to tactical detail, with multiple users connected—is essential for user trust. In defense and emergency contexts, if the visualization stutters or lags, the system is not usable. The feedback we’ve received confirms that the 3D view is responsive and dependable,” Sanchez said.

XRF’s users report clearer briefings, faster alignment across teams, more confident decisions, and smoother transitions because of the connected, integrated approach, with the 3D geospatial foundation as a key enabler.

The volcanic eruptions on the Spanish island of La Palma in 2021 were the tipping point for XRF’s founding team. Pictured here is La Palma in XRF's Apolo, a mobile app for units deployed on the ground, built with CesiumJS. Courtesy XRF.

The volcanic eruptions on the Spanish island of La Palma in 2021 were the tipping point for XRF’s founding team. Pictured here is La Palma in XRF's Colossus, a large multitouch command table for fixed control rooms, using CesiumJS. Courtesy XRF.

XRF estimates that using Cesium reduced its time to first operational deployments by 30-40%. Because the team does not need to build or maintain the core 3D terrain or streaming architecture, they are able to focus on additional features—and lower their ongoing costs by replacing disconnected tools with one shared system, just as they’re doing for their users in defense, security, and emergency management.

Learn more about optimizing, hosting, and streaming your data in an immersive, shared source of truth for time-sensitive action.

Hear from XRF at the 2026 Cesium Developer Conference. Register to join us in Philadelphia in June.

XRF Connects and Integrates Missions with Cesium

Cesiumで可視化する軌道上AIと衛星ネットワーク ―― NTTが探る「宇宙データセンタ」アーキテクチャ

Cesium Releases in June 2026

Cesium Releases in May 2026

Introducing 3D Gaussian Splats with Hierarchical Level of Detail Using 3D Tiles

XRF Connects and Integrates Missions with Cesium

Cesium Releases in April 2026

Cesium Releases in March 2026