Skyrim.AI’s Development Journey to Revolutionize Sports Media With Immersive Spatial Experiences

The Skyrim.AI Atlas platform represents the next generation of fully software-defined spatial media technology stack. Unlike traditional hardware-dependent broadcast workflows, this platform enables dynamic, real-time customization at every stage of production. By utilizing virtual "cameras" and advanced 3D capture technology, the Atlas platform ensures no moment of the game is missed. It supports both broadcast feeds and in-venue displays, providing broadcasters and stadium operators with unprecedented flexibility and efficiency.

Skyrim.AI faced several challenges in developing its spatial media platform. The lack of interoperability standards in spatial media required a collaborative approach to ensure seamless integration across production and distribution pipelines. Achieving real-time, broadcast-quality 3D reconstruction for live sports events remains a significant technological challenge. Managing the immense data load generated by volumetric capture of multiple players necessitates advanced compression techniques. Ensuring spatial media meets broadcast-quality standards is critical for adoption. Existing OTT and social media platforms require significant updates to handle the complexity of spatial media. There is a need for widespread 5G access to support high-data-rate streaming for mobile devices and XR headsets. The platform must be flexible enough to support future enhancements and adaptations to meet diverse fan needs.

NVIDIA's expertise in AI and GPU technologies has been instrumental in addressing these challenges. By utilizing tools like NIM microservices for interoperability and standards support, Skyrim.AI has built a scalable, high-performance infrastructure. NVIDIA's specialized GPUs and technologies, like NVIDIA Quantum InfiniBand networking and NVLink, ensure optimized performance across the software-defined stack. Their experience in camera-to-GPU data pipelines has enabled real-time, broadcast-quality spatial media integration.

Skyrim.AI employs a range of NVIDIA technologies to optimize its spatial media platform. NVIDIA® GPUDirect® enables efficient data transfer from cameras to GPUs, bypassing the CPU.

For training their spatial media model, Skyrim.AI utilized A100 80 GB GPUs in the cloud. For camera-to-GPU capture, they employed NVIDIA RTX™ 6000 Ada Generation GPUs, each handling four cameras. In their work with NeRF, they primarily used RTX 6000 Ada GPUs but have tested and plan to migrate to NVIDIA L40s for 3D rendering. For NIM inferencing, they currently use RTX 6000 Ada GPUs, with a transition to L40s planned. Skyrim.AI’s long-term strategy involves training on 100 series GPUs, such as H100s. The processing from cameras and the rest of the pipeline will be managed by L40s, and the AI Agent will run on lighter GPUs.

Data management technologies reduce massive data streams into manageable volumes for processing. These include tools for background removal, segmentation, and object identification, implemented through NVIDIA’s libraries, including GPUDirect and NVIDIA Instant NeRFs and Maxine NIM microservices. Proprietary AI and radiance fields are utilized for high-quality 3D reconstruction. GPU clustering allows scalable processing across multiple GPUs for large-scale events. NVIDIA NVLink and NVSwitch ensure high-bandwidth interconnection between GPUs. NVIDIA Quantum InfiniBand facilitates data transfer between GPUs across devices, ensuring scalability and reliability in distributed environments. NVIDIA Rivermax enhances the capture pipeline for efficient on-premises and cloud-based processing. It enables PTP sync at the camera, guaranteeing very tight shutter alignment, and utilizes ST 2110 low latency to get the sensor data into the GPU for processing.

The integration of NVIDIA technologies has had a transformative impact on Skyrim.AI's development. By leveraging standardized solutions like Rivermax, development time has been significantly reduced by a full year, allowing the company to focus on innovation rather than building foundational infrastructure from scratch. These technologies have also led to cost reduction, as a software-defined system eliminates the need for specialized hardware, reducing capital expenditure and venue costs. This approach empowers broadcasters with more creative tools, enabling richer content production without hardware constraints. Using appropriate GPUs for specific tasks, including NVIDIA L40s and H100, accelerates time to market and ensures efficient performance.

Solving these challenges and realizing Skyrim.AI's vision will have profound implications for the sports media industry. New revenue streams will be introduced through personalized feeds and interactive experiences. Enhanced fan engagement and immersive content will attract wider audiences and deepen fan loyalty.  The potential for expanding global fan bases is significant, as sports content tailored to diverse demographics will foster deeper connections.

Skyrim.AI's innovative spatial media platform, powered by NVIDIA technologies, is set to revolutionize the sports viewing experience. By addressing key challenges and leveraging advanced AI and GPU capabilities, Skyrim.AI is poised to redefine sports engagement, creating an inclusive and immersive future for fans worldwide.