Nvidia Unveils a New AI Tool to Help Newsrooms Detect Synthetic Video Amid Rising Deepfake Concerns

Nvidia has launched a new artificial intelligence tool designed to help newsrooms identify whether videos submitted to them were generated synthetically or captured from real-world events, part of a broader effort to preserve public trust in video journalism as AI-generated content becomes increasingly difficult to distinguish from authentic footage.

The company unveiled the tool, called Synthetic Video Detector, at Siggraph, the graphics and computer technology conference, as part of its broader Nvidia AI for Media platform. The tool is delivered as an Nvidia NIM microservice, giving media organizations a way to integrate AI-assisted video verification directly into their existing editorial workflows.

Why the tool matters

Nvidia framed the launch around the growing challenge facing journalists and broadcasters as AI video generation technology has rapidly improved. In a presentation at the conference, the company said trustworthy video remains essential to how people understand breaking news and stay connected to events unfolding around the world, given that video continues to serve as the primary medium through which audiences witness major stories in real time.

Rev Lebaredian, vice president of physical AI simulation technology at Nvidia, addressed the underlying tension driving the tool's development during a press briefing. "If we cannot tell the difference between a synthetic video and a real one, it can erode public trust when videos are presented as news, as something that came from the real world," Lebaredian said. He noted that the same underlying AI technologies capable of generating increasingly convincing synthetic videos can also be applied in reverse, to help detect when footage has been artificially created. "Fortunately, the same AI technologies we're developing to generate these videos can also be used to detect when a video is real or synthetic," Lebaredian said.

How the detection tool works

The Synthetic Video Detector analyzes submitted video on a frame-by-frame basis, producing a classifier score indicating the likelihood that the footage contains synthetic content. According to Nvidia, editorial teams can use that score to prioritize which clips require closer review, flag or quarantine questionable footage, or escalate specific videos for deeper investigation, rather than treating the score as a definitive final verdict.

Nvidia emphasized that the tool is intended to supplement, rather than replace, existing journalistic verification practices. The company noted that the microservice is specifically designed to remain effective even after videos undergo common newsroom and social media processing steps, including compression, resizing, cropping and re-encoding, all of which can otherwise degrade a detection system's accuracy.

Performance and technical specifications

In Nvidia's internal testing, the detection model achieved accuracy of up to 92% when analyzing uncompressed video, with that figure dropping to 87% at 15% compression and 82% at 50% compression, reflecting the added difficulty of identifying synthetic content in more heavily processed footage typical of real-world social media distribution.

The system is capable of processing 1080p video in as little as 22 milliseconds on Nvidia RTX systems, and approximately 30 milliseconds on Nvidia L40 GPUs, according to the company. Nvidia said the latest revision of the underlying model has also demonstrated improved internal benchmark performance, achieving an area-under-the-curve, or AUC, score of 0.9614 and an accuracy rating of 0.9453 on Nvidia's internal test set. The AUC metric measures how effectively a classifier system ranks genuinely synthetic content above authentic footage, independent of any specific detection threshold. Nvidia said the system's thresholds can be adjusted to accommodate different review postures, including more conservative configurations designed to reduce the risk of synthetic video going undetected.

A tool built for sensitive deployment environments

Nvidia designed the microservice to be deployable in a range of environments depending on an organization's specific security and operational needs, including on-premises systems, edge computing setups, hybrid infrastructure, and approved air-gapped environments that remain isolated from external networks. That flexibility is intended to help organizations maintain tighter control over sensitive video data, access permissions and overall operational security.

Early industry adoption

Streaming infrastructure company Wowza has already begun integrating the Synthetic Video Detector through its Wowza Video Intelligence Framework, bringing real-time synthetic video detection capabilities into livestreaming workflows used across more than 35,000 deployments spanning over 170 countries. Nvidia noted that this kind of scaled adoption is particularly significant given that many organizations most exposed to synthetic media risks, including broadcasters, government agencies, financial institutions and operators of critical infrastructure, also face strict regulatory requirements around data residency, security and operational control.

By pairing the detection tool with video infrastructure that customers already use, Wowza aims to make AI-assisted verification available closer to the point where video is initially captured and streamed, allowing organizations to flag potentially synthetic content in real time while keeping sensitive footage contained within their own secured environments.

What the tool does not do

Nvidia was clear in clarifying the specific scope and limitations of the detection service. When asked how the tool determines whether a video is fake, the company said the system does not evaluate whether the events depicted in a video are factually accurate or true. Instead, it strictly assesses whether the video itself was synthetically generated, using dual vision-transformer models known as DINOv2 and DINOv3 to analyze individual frames, producing frame-level classification scores that are then averaged and combined into an overall video-level synthetic-likelihood score.

The model was trained and evaluated using a combination of real and synthetic video content, including footage generated through diffusion-based AI methods. Nvidia reiterated that newsrooms should treat the resulting scores as a triage and escalation tool rather than a definitive final judgment, emphasizing that human editorial oversight and established verification practices remain essential components of the overall process.

With Synthetic Video Detector now available and early adoption already underway through partners like Wowza, Nvidia's broader push into AI-assisted media verification tools is likely to continue expanding as concerns about synthetic and AI-generated content grow across the journalism and broadcasting industries. As AI video generation technology continues to advance in realism and accessibility, tools like Nvidia's detector are expected to become an increasingly important, though not standalone, part of how newsrooms work to preserve public trust in the video content they publish and distribute.