Google has expanded its Gemini artificial intelligence video verification function made or edited using the company’s own artificial intelligence models. Users can now ask Gemini to determine whether an uploaded video was generated by AI by asking: “Was it generated using Google AI?”
Gemini will scan the video’s graphics and audio for a proprietary Google watermark called SynthID. Google says the answer will be more than just “yes” or “no.” Gemini will indicate specific times when a watermark appears in a video or audio. The company rolled out the feature for images in November, also limiting it to images created or edited with Google AI.
Some watermarks may be easy to scrubas OpenAI found out by running the Sora app full of videos generated exclusively by artificial intelligence. Google calls its own watermark “imperceptible.” That said, we don’t yet know how uncomplicated it will be to remove them, or how easily other platforms will detect SynthID information and mark the content as AI-generated. Google’s Nano Banano AI image generation model in Gemini includes C2PA metadata, but a general lack of coordinated tagging of AI-generated material across social media platforms allows deepfakes to go undetected.
Gemini can handle videos up to 100 MB in size and 90 seconds in length for verification. This feature is available in every language and location where the Gemini app is available.