Gemini App from Google Now Analyzes Videos for AI Creation

Google has broadened the scope of Gemini's AI inspection capabilities to include videos created or modified using their AI technologies. This feature allows users to query the app with, “Did Google AI generate this video?”

Gemini scrutinizes both the visual and auditory content of the video to detect a distinctive mark known as SynthID, which is embedded by Google. The tool goes beyond merely confirming or denying AI involvement; it provides precise timestamps indicating when the watermark surfaces in the content.

This functionality was initially launched for images in November, but it is restricted to content crafted or altered with Google’s AI systems.

Understanding SynthID's Role

While some conventional watermarks are easy to erase—a fact underscored by OpenAI's experience with its Sora app featuring AI-only videos—Google claims that its own watermark is 'undetectable to the human eye.' The question remains whether SynthID will be simple to eliminate or if it will be efficiently recognized by other digital platforms for labeling AI-driven content.

The embedded identifier is part of Google's image creation system integrated within Gemini, yet the proliferation of AI-generated works online often lets digitally altered media evade detection.

Limits and Language Support

Gemini's verification option is capable of analyzing videos of up to 100 MB in size and lasting no more than 90 seconds. This service is available wherever the Gemini app itself is deployed and supports all languages in those regions.