NVIDIA said at IBC 2026 in Amsterdam that it is significantly expanding its NVIDIA AI for Media suite, adding tools including GPU-accelerated SDKs and NIM microservices for broadcast, sports, and streaming workflows. The company said the suite is designed to bring real-time AI capabilities to media operations, covering motion understanding, video verification and enhancement, program localization, and the development of AI-driven media applications. A key release is the NVIDIA Synthetic Video Detector, or SVD, offered as a NIM microservice that helps organizations assess the likelihood that a video clip is authentic or AI-generated. NVIDIA said the model reaches 99.3% accuracy for text-to-video content and 97.7% for image-to-video content. The company also said the technology is being integrated into products from Dalet, a Dassault Systèmes subsidiary, video intelligence platform TwelveLabs, and streaming provider Wowza. In addition, NVIDIA presented its 3D Body Pose technology, which estimates 2D and 3D human joint positions and angles from single-camera video and converts movement into structured data for uses including sports analysis and content creation. Vizrt is already using the technology in real-time virtual studio environments, according to NVIDIA Blog.

ByteDance is developing a real-time spatial video generation AI model for robotics and autonomous systems, competing with Meta and Alphabet. Founder Zhang Yiming is personally supervising, coordinating resources across departments. A release could come as early as next month.

Elon Musk said xAI’s image-generation tool Grok Imagine has reached positive gross profit and stressed that video understanding and generation will be critical to advancing AGI.