As artificial intelligence makes it easier than ever to generate convincing fake videos, news organizations are facing a growing challenge: determining whether footage is authentic before publishing it.
To address that problem, NVIDIA on Monday unveiled Synthetic Video Detector, a new AI-powered tool designed to help journalists and media organizations identify videos that may have been generated or manipulated using AI. The company announced the technology at the SIGGRAPH conference in Los Angeles as part of its NVIDIA AI for Media platform.
NVIDIA unveiled Synthetic Video Detector, a new AI-powered tool designed to help identify videos that may have been generated or manipulated using AI
(Photo: Courtesy of NVIDIA)
The launch comes at a time when generative AI has dramatically lowered the barrier to creating so-called deepfakes — realistic synthetic videos that can depict people saying or doing things that never happened. Such content has become increasingly common online, raising concerns about misinformation, election interference and the spread of manipulated footage during breaking news events.
Unlike traditional deepfake detection tools that attempt to deliver a simple yes-or-no answer, NVIDIA's system is designed to support existing newsroom verification processes. The software analyzes video frame by frame and produces a probability score estimating whether footage contains synthetic content. Editors can then use that score to prioritize videos for review, flag suspicious material or escalate it for more detailed forensic analysis.
The company says the tool is intended to provide an additional signal for journalists working under tight deadlines rather than replace established editorial verification methods.
"Public trust in video is more essential than ever," NVIDIA said in announcing the new technology, adding that the detector is designed to help newsrooms make faster decisions while maintaining editorial standards.
One of the biggest technical challenges in detecting manipulated video is that clips are often altered after they are created. Videos uploaded to social media platforms are routinely compressed, resized, cropped and re-encoded, processes that can remove many of the digital clues forensic tools rely on.
According to NVIDIA, Synthetic Video Detector is designed to remain effective despite those changes. In internal testing, the company said the model achieved up to 92% accuracy on uncompressed video, 87% accuracy after 15% compression and 82% accuracy after 50% compression. While those figures have not been independently verified, they suggest the tool can maintain much of its effectiveness even after videos have undergone significant processing.
The company also says the system is fast enough for real-world newsroom workflows. It can analyze 1080p video in as little as 22 milliseconds on NVIDIA RTX systems and approximately 30 milliseconds on NVIDIA L40 graphics processors, enabling near real-time analysis.
NVIDIA is positioning the software not only for media organizations but also for government agencies, financial institutions and operators of critical infrastructure that increasingly face the risk of AI-generated misinformation. The detector can be deployed in on-premises, hybrid, edge or air-gapped environments, allowing organizations to keep sensitive video within their own infrastructure rather than sending it to cloud services.
The technology is already beginning to see commercial adoption. Live-streaming technology provider Wowza has integrated Synthetic Video Detector into its Wowza Video Intelligence Framework, enabling customers to screen livestreams for potentially synthetic content in real time. NVIDIA said Wowza's platform is used in more than 35,000 deployments across over 170 countries.
The announcement reflects a broader trend as technology companies race to develop tools capable of identifying AI-generated media even as generative models continue to improve. While image and video generators have become significantly more sophisticated over the past two years, experts caution that no detection system is foolproof. Many news organizations continue to rely on a combination of digital forensic tools, metadata analysis, geolocation techniques and traditional reporting to verify visual content.
Rather than serving as a definitive judge of authenticity, NVIDIA's new detector is intended to become another tool in that broader verification toolkit, helping journalists decide which videos deserve closer scrutiny before they reach the public.