NVIDIA's Synthetic Video Detector can detect AI-generated video with 92% accuracy in less than a second

Q: What accuracy does NVIDIA’s Synthetic Video Detector NIM microservice achieve on uncompressed and compressed video?

NVIDIA reports the Synthetic Video Detector NIM microservice achieves 92% detection accuracy on uncompressed video. Its accuracy drops to 85% for 15% compression and 82% for 50% compression.

Q: How fast can the Synthetic Video Detector process 1080p video on consumer RTX and Ada Lovelace L40 GPUs?

NVIDIA says the Synthetic Video Detector can process 1080p video in as little as 22 milliseconds on an RTX system (GPU unspecified), and around 30 milliseconds on an Ada Lovelace NVIDIA L40 enterprise-grade data center GPU.

Q: How does video compression level (15% and 50%) affect the detector’s accuracy?

NVIDIA reports the detector reaches 92% accuracy on uncompressed video, which falls to 85% with 15% compression and to 82% with 50% compression. Thus higher compression reduces the detector's accuracy, with a 7 percentage point drop at 15% compression and a 10 point drop at 50% compression.

Q: Can the NIM microservice be integrated into livestreaming workflows and at scale?

Yes. NVIDIA says the Synthetic Video Detector NIM microservice is designed to scale and integrate with systems and "can also handle synthetic video detection in livestreaming workflows." NVIDIA also notes it has been deployed at scale by Wowza through its Wowza Video Intelligence Framework, which is used in over 35,000 locations in over 170 countries.

The new Synthetic Video Detector NVIDIA NIM microservice can be integrated at scale to detect AI-generated video content with an accuracy of up to 92%.

VIEW GALLERY - 2

Kosta Andreadis

Senior Editor

Published Jul 21, 2026 2:02 AM CDT

1 minute & 15 seconds read time

TL;DR: NVIDIA's Synthetic Video Detector NIM microservice analyzes video frame-by-frame to flag AI-generated content, achieving 92% accuracy on uncompressed video (85% at 15% compression, 82% at 50%). It processes 1080p video in ~22-30 ms on NVIDIA GPUs, scales for livestreams, and integrates with newsroom workflows.

Voice: Kosta AndreadisSpeed

0:00 / 3:03

In the generative AI era, if you're on social media, then odds are you've been seeing or watching several AI-generated videos. And with the lifelike quality of generative AI content now at a point where it can be hard to tell what's "real" or not, there's a genuine need for AI content to not only be flagged but detected to ensure transparency and minimize the spread of misinformation.

NVIDIA's Synthetic Video Detector can detect AI-generated video with 92% accuracy in less than a second 2

VIEW GALLERY - 2 IMAGES

At the latest SIGGRAPH event that's currently underway in Los Angeles, NVIDIA announced the Synthetic Video Detector NVIDIA NIM microservice, which is built for media outlets, newsrooms, and enthusiasts. It's an AI-assisted detection model designed to detect AI-generated video content.

According to NVIDIA, this NIM microservice "analyzes video frame by frame to produce a classifier score of whether it contains synthetic content." And with that, individuals or teams can then review and analyze flagged clips and videos. NVIDIA notes that based on its own internal testing, the Synthetic Video Detector NVIDIA NIM microservice model's detection accuracy reached 92% on uncompressed video.

Read more: YouTube is making its AI Deepfake detection tool available to all creators over the age of 18

This figure drops to 85% for 15% compression video and 82% for 50% compression, so even in the digital age of codecs and artifacts, it can still accurately detect AI-generated video content. The upside for media and newsrooms is that it can process 1080p video in as little as 22 milliseconds on an RTX system (GPU unspecified), or around 30 milliseconds on an Ada Lovelace NVIDIA L40 enterprise-grade data center GPU.

It's also designed to scale and integrate with systems, with NVIDIA noting that it can also handle synthetic video detection in livestreaming workflows. One company that has already deployed it at scale is Wowza, whose Wowza Video Intelligence Framework is deployed in over 35,000 locations in over 170 countries.

Frequently Asked Questions

TweakBot answers common questions about this news using TweakTown's own coverage from this page and related content from our archive. Tap a question to reveal the answer, or type your own below.

Question #1

What accuracy does NVIDIA’s Synthetic Video Detector NIM microservice achieve on uncompressed and compressed video?

Question #2

How fast can the Synthetic Video Detector process 1080p video on consumer RTX and Ada Lovelace L40 GPUs?

Question #3

How does video compression level (15% and 50%) affect the detector’s accuracy?

Question #4

Can the NIM microservice be integrated into livestreaming workflows and at scale?

Have a question not listed here? Ask below and TweakBot will answer it.

"That scale matters because many of the organizations most exposed to synthetic media risk, including broadcasters, government agencies, financial institutions and critical infrastructure operators, also face strict requirements around data residency, security and operational control," NVIDIA writes. "Rather than replacing established verification practices, the microservice provides another signal for time-sensitive decisions - helping teams move quickly while protecting editorial standards and ensuring public trust."