[Digital Today reporter Jae-won Choi (최재원)] Nvidia has significantly expanded Nvidia AI for Media, an AI tool suite for broadcasting, sports coverage and global streaming. The focus is to handle everything from video frame generation and high-resolution conversion to athlete motion tracking, real-time lip sync and voice enhancement in a single AI pipeline.

Nvidia said on Sept. 15 that its new Video Frame Generation (VFG) creates AI-generated intermediate images between existing frames to raise the frame rate by 2 times or 4 times. In sports broadcasting, it can be used for smoother replays and slow-motion production. Generation at 6 times and 8 times is still in the experimental and evaluation stage.

Video Super Resolution (VSR) upscales low-resolution video up to 8K while reducing noise, blur and compression artifacts. It also supports 10-bit video. TrueHDR converts SDR video into HDR in real time, keeping the original contrast while expanding brightness representation up to about 2,000 nits. VFG, VSR and TrueHDR can be applied sequentially in a single video effects pipeline.

The company also strengthened its person analysis functions. 3D Body Pose tracks full-body joint movement from single-camera video without separate markers and converts it into data for 3D animation. It can be used for sports athlete analysis, replays and animation production.

For global content localisation, Nvidia uses LipSync and Active Speaker Detection. LipSync adjusts a person’s mouth movements on video in real time to match translated audio, and Active Speaker Detection analyses video and audio together to identify the actual speaker among multiple people. It also supports lip sync in French, German and Spanish.

In audio, Studio Voice corrects ordinary microphone audio to broadcast-quality, and a separate noise removal function reduces ambient noise. Nvidia also improved its existing AI-generated video detection tool, Synthetic Video Detector. Its strategy to apply AI across the full broadcasting production process, from generation to editing, translation and verification, appears to be taking more concrete shape.