How does Google Gemini verify AI videos?

Gemini scans videos and audio for Google's invisible SynthID watermark to detect AI generation.

Can Gemini detect deepfake videos?

Gemini's new feature aims to help identify AI-generated videos, potentially including deepfakes.

What are the limitations of Gemini's video verification?

Gemini can verify videos up to 100 MB and 90 seconds in length.

Home / Technology / Google's Gemini: New Tool Detects AI-Generated Videos

Google's Gemini: New Tool Detects AI-Generated Videos

19 Dec

•

Summary

Gemini now verifies AI-generated videos using SynthID watermark.
The feature scans video and audio for Google's proprietary watermark.
Gemini handles videos up to 100 MB and 90 seconds for verification.

Google's Gemini: New Tool Detects AI-Generated Videos

Google's Gemini AI is enhancing its verification capabilities by extending AI-generated content detection to videos. Users can now query Gemini, asking "Was this generated using Google AI?" to ascertain if a video was created or modified by Google's AI models.

Gemini performs a scan of both the visual and audio elements within a video, searching for Google's proprietary SynthID watermark. Unlike a simple yes or no, Gemini will pinpoint specific instances where the watermark is present. This feature is accessible for videos up to 100 megabytes and 90 seconds in length.

This development follows the introduction of a similar image verification tool in November. While Google describes its SynthID watermark as "imperceptible," the broader challenge of coordinated AI content tagging across social media platforms remains, posing ongoing risks for undetected deepfakes.