Google’s Gemini app can check videos to see if they were made with Google AI

Google expanded Gemini’s AI verification characteristic to videos made aliases edited pinch the company’s ain AI models. Users tin now inquire Gemini to find if an uploaded video is AI-generated by asking, “Was this generated utilizing Google AI?”

Gemini will scan the video’s visuals and audio for Google’s proprietary watermark called SynthID. The consequence will beryllium much than a yes aliases no, Google says. Gemini will constituent retired circumstantial times erstwhile the watermark appears successful the video aliases audio. The institution rolled retired this capability for images successful November, besides constricted to images made aliases edited pinch Google AI.

Some watermarks tin beryllium easily scrubbed, arsenic OpenAI learned erstwhile it launched its Sora app afloat of exclusively AI-generated videos. Google calls its ain watermark “imperceptible.” Still, we don’t yet cognize really easy it will beryllium to remove, aliases really readily different platforms will observe the SynthID accusation and tag the contented arsenic AI-generated. Google’s Nano Banano AI image procreation exemplary wrong the Gemini embeds C2PA metadata, but the wide lack of coordinated tagging of AI-generated worldly crossed societal media platforms allows deepfakes to spell undetected.

Gemini tin grip videos up to 100 MB and 90 seconds for verification. The characteristic is disposable successful each connection and location that the Gemini app is available.