Spatial computing is moving from visualization to lively collaboration, adding increasingly more GPU demands on XR hardware to render photorealistic, physics-accurate, high-fidelity spatial content in real time. Meanwhile, developers have had to keep up separate codebases for each platform, each with different toolchains, SDKs, and streaming protocols.
At NVIDIA GTC 2026, NVIDIA CloudXR 6.0 introduced a universal OpenXR-based streaming runtime that works across headsets, operating systems, and browsers—including native visionOS integration. This post walks through how the CloudXR 6.0 architecture works and tips on how to start constructing today.
CloudXR 6.0: Universal OpenXR streaming
The discharge focuses on expanding the reach of NVIDIA RTX-powered content to any spatial display without the constraints of local hardware or manual device provisioning.
Native spatial streaming for Apple platforms
NVIDIA and Apple have collaborated to construct a high-performance bridge for Apple Vision Pro using privacy-protected foveated streaming enabled by visionOS 26.4. With visionOS for CloudXR, developers can stream high-fidelity, high-frame-rate, and low-latency graphically demanding PC simulations and skilled 3D applications—like Autodesk VRED, iRacing, and X-Plane—on to Apple Vision Pro.
The transition from standard high-resolution streaming to dynamic foveated streaming represents a very important step forward in how spatial content is delivered. In standard streaming architectures, the system must encode and transmit a full-size frame at 4K resolution to make sure visual clarity across the complete field of view. This approach is resource-heavy and infrequently pushes the bounds of network bandwidth and client-side processing.
Dynamic foveated streaming changes the paradigm by optimizing content based on roughly where a user is looking. As an alternative of a uniform 4K stream, the system transmits at 1K resolution while maintaining the identical pixel density on the fovea (the middle of the user’s focus). Despite the huge reduction in data, the system maintains a visible quality that’s perceptually just like a full 4K stream since the high-resolution detail is at all times where the user is looking.
By adopting foveated streaming, developers and enterprises can unlock 4 critical advantages:
- High-quality wireless freedom: Stream 4K-like spatial experience and 90 FPS reliably over standard 5GHz WiFi networks.
- Scalable multi-user environments: The reduced bandwidth per user enables simultaneous streaming for multiple users on the identical network infrastructure— essential for collaborative design.
- Ultra-low latency: Performance is boosted by reduced transmission and faster encoding and decoding, minimizing motion-to-photon lag.
- Hardware efficiency and battery life: For devices like Apple Vision Pro, this directly translates to preserved battery life for longer sessions in the sector or the office.
The combination between visionOS and NVIDIA CloudXR is anchored in a privacy and security by design philosophy. Developers can profit from foveated streaming without accessing sensitive focus region data. This ensures that the user’s gaze stays private while the applying gains the obligatory performance optimizations.
Xcode offers a foveated streaming app template that allows the creation of a completely functional, multi-platform application inside seconds. Developers can get a pre-configured foundation for spatial streaming with only a couple of clicks.
Visit the developer and documentation pages for more information.
CloudXR.js: Zero-friction web access
CloudXR.js provides a zero-install path for developers, enabling access to robotics, NVIDIA Omniverse, and OpenXR content through a single web link. By utilizing WebRTC and WebXR, it removes the necessity for native application installs on devices like Meta Quest 3 and PICO 4 Ultra.
Getting began with CloudXR 6.0
The CloudXR 6.0 Runtime SDK is distributed as a set of shared libraries and C headers. The really useful integration path is direct SDK integration: linking the CloudXR libraries into your application and managing the service lifecycle programmatically on each Windows and Linux servers. This offers your application full control over when and the way streaming runs.
Server prerequisites
- NVIDIA RTX GPU (RTX 6000 Ada Generation or higher really useful)
- Windows 11 (64-bit) or Ubuntu 22.04+ (64-bit)
- An existing OpenXR application
- CloudXR 6.0 Runtime SDK, downloaded from NVIDIA NGC
Network prerequisites
- IPv4 connectivity
- 200 Mbps minimum for local streaming
Integrating the CloudXR Runtime right into a streaming application
The CloudXR Runtime SDK ships with two integration surfaces: the OpenXR runtime manifest openxr_cloudxr.json tells the OpenXR loader to route your application’s XR sessions through CloudXR, and cxrServiceAPI.h, a C API for managing the complete service lifecycle.
For an entire reference implementation, the CloudXR LÖVR sample shows full lifecycle management and OpenXR session establishment with clients. Developers working in Unreal Engine can follow the dedicated guide.
High-level steps of integrating CloudXR Runtime right into a streaming application:
Step 1
- Download the CloudXR 6.0 Runtime from the NVIDIA NGC catalog. The package includes the shared libraries, the OpenXR runtime manifest, and the combination headers needed to embed CloudXR into your application.
Step 2
- Point the OpenXR loader to the
openxr_cloudxr.jsonmanifest file included within the SDK. - This registers CloudXR because the lively runtime in your server for Windows and Linux in order that the OpenXR application routes its sessions through CloudXR mechanically.
- Once registered, no changes to the applying’s rendering code are required. See the CloudXR Runtime documentation for platform-specific registration steps.
Step 3
- Link your application against the CloudXR service library and include
cxrServiceAPI.h. This offers your application programmatic control over the CloudXR service—when it starts, the way it is configured, and when it stops—with the service running as a component alongside your OpenXR application.
Step 4
- Use
nv_cxr_service_create()to instantiate the service object, then configure it using the property-setting functions in cxrServiceAPI.h before calling start. - Properties cover string, boolean, and numeric parameters, including options like bitrate targets. Check with the CloudXR Runtime documentation for the complete list of supported property names and values.
Step 5
- Call
nv_cxr_service_start()to bring the service online. - Once running, poll the event queue in your application loop to trace connection state—the service surfaces events for each your OpenXR application and streaming clients as they connect and disconnect.
- This lets your application respond intelligently to session changes, reminiscent of pausing rendering when no client is present or logging lifecycle events for diagnostics.
Step 6
- When shutting down, signal the service to stop, wait for it to totally terminate, then release the service object.
- The CloudXR Runtime API reference covers the complete shutdown sequence.
Step 7
- For applications that must exchange custom data alongside the video stream—telemetry, simulation state, or commands beyond standard OpenXR actions—CloudXR supports the
XR_NV_opaque_data_channelextension. - This provides a bidirectional byte channel between server and client, identified by a shared UUID.
- See the CloudXR Opaque Data Channel documentation for integration details.
Alternative to integrating: Stream Manager (Windows only)
For Windows deployments where decoupling runtime management from the applying process makes more sense, the CloudXR SDK also includes Stream Manager. This standalone Windows service manages CloudXR runtime instances through an RPC interface. This is especially useful when multiple applications must share a single streaming service.
Stream Manager is required to work with the Foveated Streaming in visionOS. See the CloudXR documentation for setup and configuration.
Select a client integration path
With the server-side integration complete, select which client SDK to integrate based in your goal platform.
Apple platforms (visionOS, iOS, and iPadOS)
The CloudXR framework is distributed as a native Swift Package. Add CloudXRKit as a dependency in your Xcode project, initialize a streaming session pointed at your server, and the framework handles hardware-accelerated video decoding, spatial tracking via ARKit, and native scene rendering through RealityKit while streaming RTX-rendered frames.
Dynamic foveated streaming—using Apple’s FoveatedStreaming API—delivers high-resolution content to the world where the user is looking while keeping gaze data protected, reducing bandwidth requirements without sacrificing perceived quality. For developers ranging from scratch, CloudXR also provides an Xcode project template that bootstraps an Apple Vision Pro, iPhone, or iPad client with no boilerplate required.
For a full technical walkthrough, see spatial streaming for Apple platforms and the Apple Platforms developer page.
Web browsers and standalone headsets (CloudXR.js)
CloudXR.js is a JavaScript library that allows streaming on to WebXR-capable devices—Meta Quest 3, Pico 4 Ultra, and desktop browsers—through a single web link. It uses WebRTC for low-latency video delivery and the WebXR Device API for display and input, and is compatible with WebGL, Three.js, and React Three Fiber.
For more details, see the CloudXR.js developer page.
Start
For developers able to start, integrate the CloudXR Runtime SDK into your existing OpenXR application, register the runtime, and select the client SDK that matches your audience. The identical server-side integration supports every client platform CloudXR offers.
Resources:
