Nvidia has introduced a new tool called the Synthetic Video Detector. This is a NIM microservice built to help news and editorial teams identify videos that have been generated or modified using artificial intelligence. While AI offers many benefits for creativity, it also increases the spread of fake news and raises concerns about public trust.
The new detector examines video content frame by frame and generates a classification score that indicates whether a clip is likely to have been created or modified by AI. This tool does not replace traditional validation methods, but rather serves as a fast first layer of screening. This allows editors to quickly flag suspicious clips, isolate questionable footage, or send material for further analysis before it’s published or broadcast.
Nvidia unveils synthetic video detector for AI-generated content
One of Synthetic Video Detector’s main strengths is its balance between speed and accuracy. Nvidia says the tool can achieve up to 92% accuracy when analyzing uncompressed video. Even when compressing videos that often lose detail, light compression maintains 87% accuracy and high compression maintains 82% accuracy.
This system is designed for high speed processing. A 1080p video can be analyzed in about 22ms when running on a local Nvidia RTX system, while processing takes about 30ms on an Nvidia L40 server GPU.
Another feature is the ability to adjust the detection threshold. This allows news organizations to set stricter detection levels if their top priority is to prevent AI-generated videos from appearing on air.
To bring this technology to more users, Nvidia partnered with Wowza to integrate microservices into its intelligent video platform. The company says the detector will soon be available in more than 35,000 deployments around the world and will help reduce the likelihood that fake AI-generated videos will be aired by news organizations.
