Nvidia is bringing automated video forensics into the newsroom with a new detection tool that aims to help media organizations separate authentic footage from increasingly convincing AI-generated fabrications.
A microservice built for broadcast accuracy
The company introduced its Synthetic Video Detector during the SIGGRAPH 2026 conference. Delivered as an NIM microservice, the system inspects each frame of a video—whether a file or a live feed—and generates a predictive score indicating the likelihood that the content was produced by generative AI. The offering is designed to support editorial teams in verifying material rapidly and flagging synthetic media before it reaches broadcast.
Rev Lebaredian, Nvidia’s vice president of physical AI simulation technology, said the core detection logic draws on the same underlying techniques used to create generative video. In practice, this means media organizations can deploy a detection layer that evolves alongside the generative models it is meant to catch.
Speed and accuracy under real-world conditions
The tool’s performance varies depending on video quality. On uncompressed footage, Nvidia reports a detection accuracy of 92 percent. That figure decreases to 87 percent when the video undergoes 15 percent compression, and settles at 82 percent under 50 percent compression—a drop the company describes as relatively moderate given the heavy data loss.
Processing speed is central to live-production workflows, and the microservice returns a result in 22 milliseconds per 1080p frame on RTX-class systems, or 30 milliseconds on L40 GPUs. In internal benchmarking using the most recent model version, Nvidia recorded an area-under-curve score of 0.9614 and an accuracy of 94 percent. Organizations can calibrate detection thresholds to align with their specific editorial standards and risk tolerance.
Early deployment and industry reach
Nvidia is already distributing the Synthetic Video Detector through Wowza, an enterprise-grade live video and media server platform used by numerous news networks worldwide. Wowza CEO Krish Kumar confirmed that the detector is being embedded into the company’s infrastructure, part of a broader push to apply AI processing while a live stream is still active. “The market is moving toward live video systems that can generate value while the stream is still in motion,” Kumar said.
The integration falls under Wowza’s Video Intelligence Framework, which reaches roughly 35,000 deployments across close to 170 countries. That footprint means real-time synthetic-video flagging is poised to extend well beyond traditional news studios, reaching sports broadcasters, corporate streaming platforms, and live-event producers who face the same verification challenges as AI-generated media becomes more sophisticated.
Sources: blogs.nvidia.com, www.streamingmedia.com