Release Notes#
Current Release#
26.07.0#
Release Date: July 2026
Omniverse Streaming on NVCF 26.07.0 is the first general availability (GA) release. It is a self-hosted Reference Architecture for deploying and scaling streaming applications on NVIDIA Cloud Functions (NVCF), running on AWS or Azure infrastructure and built on the NVCF v0.6.0 platform.
Highlights#
Self-hosted, self-managed deployment: Run the full NVCF control plane and GPU compute plane in your own cloud environment, operated independently, with no dependency on NGC-hosted infrastructure.
Reference Architecture: A defined set of artifacts, versions, and steps that fit together into a validated streaming deployment. Follow it as documented, or adapt it to your environment.
Multi-cloud: Deploy on Amazon Web Services (EKS and ECR) or Microsoft Azure (AKS and ACR).
Real-time interactive streaming (LLS): Low Latency Streaming carries WebRTC media and low-latency input between a browser client and a Kit application on NVCF. It uses a single configurable UDP media port behind a static ingress, with an optional TURN server that falls back to TCP or TLS for clients on networks that block the direct UDP path.
On-demand autoscaling: A built-in function autoscaler adjusts running instances with demand across clusters.
Layered caching: Container Cache (image pulls), GXCache (shader compilation), Derived Data Cache Service (DDCS), and Omniverse Content Cache (OVCC) reduce startup and scene-load time.
Storage integration: Connect your own S3 or Azure object storage as working storage through standard adapters, with real-time file-change notifications for live stage updates.
End-to-end observability: Distributed tracing across service boundaries, session-correlated logs, and metrics exported in OpenTelemetry (OTEL) format to your existing tools.
One stack, many GPU workloads: The same self-hosted stack also runs AI inference, simulation, and fine-tuning on the same clusters, so you do not maintain separate stacks.
Kit application to Cloud Function: Scaffold a streaming Kit application from the Kit App Template, add the NVCF streaming layer, containerize it, and deploy it as a Cloud Function through the NVCF CLI or API.
Web client scaffolding: Generate a ready-to-run WebRTC web client with the Create OV WebRTC App generator, the Omniverse Web Streaming Library, and the intermediary proxy.
Agentic AI Skills: AI-agent skills plan and deploy the self-hosted stack across AWS and Azure, covering infrastructure, NVCF, caches, storage, and validation. See Agentic AI Skills.
NVCF supports in-place upgrades: This release runs on the NVCF v0.6.0 platform, and existing NVCF v0.5.0 deployments can upgrade to v0.6.0 in place. For platform-level changes and upgrade steps, see the NVCF v0.6.0 release notes.
Open-source NVCF: NVCF, the platform layer beneath this Reference Architecture, is open source under Apache 2.0. Only NVCF is open source; the Reference Architecture itself is a mix of open and closed source. See the github.com/NVIDIA/nvcf repo for details.
OKAS 1.0 migration: For teams on Omniverse Kit App Streaming (OKAS) 1.0, the migration guide maps the concepts and walks through moving to this release (a fresh deployment, not an in-place upgrade).
Component Versions#
This Omniverse Streaming on NVCF 26.07.0 release is built on the NVCF v0.6.0 platform. For the Reference Architecture components, see the Artifact Manifest; for the NVCF platform artifacts, see the NVCF v0.6.0 Artifact Manifest.
Important
Omniverse Streaming consists of artifacts (Helm charts and containers) that are both publicly available and gated. While the publicly accessible components are available on NGC, to gain access to the gated components, contact your NVIDIA Account Manager or click here for more information.