Google’s Gemini App Can Verify If Videos Are Created Using Google AI

Google has enhanced the AI verification capabilities of Gemini to include videos created or modified with its AI models. Users can now inquire whether a video was generated using Google AI by asking, “Was this generated using Google AI?”

Gemini will analyze the video’s visuals and audio for a proprietary watermark known as SynthID. Instead of a simple yes or no, Gemini will provide detailed feedback, indicating specific timestamps where the watermark is present in the video or audio. This feature was previously introduced for images in November, also limited to those created or edited with Google AI.

While some watermarks can be easily removed, as demonstrated by OpenAI with its Sora app featuring solely AI-generated videos, Google claims its watermark is “imperceptible.” However, it remains unclear how easily SynthID can be eliminated or how effectively other platforms will identify and label content as AI-generated. Google’s Nano Banano AI image generation model within Gemini incorporates C2PA metadata, but the inconsistent tagging of AI-generated content across social media platforms allows deepfakes to slip through the cracks.

Gemini supports video verification for files up to 100 MB and 90 seconds in length. This feature is accessible in all languages and regions where the Gemini app is available.