NVIDIA Launches 22ms Deepfake Detector
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- NVIDIA launched Synthetic Video Detector NIM, a microservice that analyzes videos frame by frame to classify synthetic content with up to 92% accuracy on uncompressed video
- NVIDIA states the tool processes 1080p video in 22ms on RTX systems and 30ms on L40 GPUs, enabling real-time detection for time-sensitive media verification
- NVIDIA reports the model achieves an AUC of 0.9614 and 94.53% accuracy on internal benchmarks, outperforming open-source detectors on the AI GVD bench leaderboard
- Wowza will embed the Synthetic Video Detector into its Intelligence Video framework, making it available across over 35,000 deployments in 170 countries
- NVIDIA emphasizes the tool does not replace human verification but adds a scalable layer for editorial teams to flag, prioritize, or quarantine suspicious video content
Why it matters: Media organizations gain a fast, automated way to triage potentially fake video at scale, reducing the risk of spreading misinformation—especially critical during breaking news. The 22ms analysis speed on consumer-grade RTX hardware makes this deployable beyond data centers, shifting part of the verification burden from humans to AI infrastructure.
