AI Briefing
KO

NVIDIA Unveils Sports AI Fine-Tuning and Real-Time Media Processing Technologies at IBC 2026

·2026.09.10 01:00

Key point

NVIDIA unveiled an AI playbook that significantly boosts accuracy through sports video fine-tuning, along with real-time media processing technologies, at IBC 2026.

1 / 3

Details

NVIDIA announced AI technology expansions for broadcasting, sports, and global streaming at IBC 2026. The core offering is the NVIDIA Sports Intelligence Playbooks, which enable leagues and media companies to fine-tune open models using their own data. Initial test results showed multiple-choice accuracy improving from approximately 53% to 94%, and subjective evaluation scores rising from 5.7% to 66%, indicating a significant improvement in contextual understanding.

Real-Time Media Processing and Verification Technologies

The NVIDIA Synthetic Video Detector (SVD) is a NIM microservice for identifying AI-generated videos, achieving 99.3% accuracy for text-to-video and 97.7% for image-to-video. It is integrated with partners such as Dalet, TwelveLabs, and Wowza for use in news verification and compliance workflows.

Additionally, Video Frame Generation (VFG) uses generative AI to double or quadruple frame rates. Integrated into Ross Video's Rio Replay platform, it generates 6x slow motion without requiring ultra-high frame rate cameras. NVIDIA 3D Body Pose estimates human joint positions from a single camera, replacing markerless motion capture, and is used for real-time lighting effect control in Vizrt's live virtual studios.

Software-Defined Broadcasting and Localization

Holoscan for Media and the Media Exchange Layer (MXL) provide an open architecture that allows AI processing and traditional media functions to run on the same accelerated infrastructure. This enables developers to improve infrastructure efficiency and rapidly deploy new features.

Content Localization technology is integrated into the Holoscan for Media toolkit to process subtitles, translated audio, dubbing, and synchronized video in real time. Since only the necessary features can be selected per program, there is no need to build separate infrastructure for each localized version. Improvements to the LipSync and Active Speaker Detection NIMs have enhanced the quality and naturalness of multilingual broadcasts.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.