Google has expanded the capabilities of its Gemini app by introducing an AI verification feature for videos created or edited using its own AI models. Users can now leverage Gemini to determine if an uploaded video was AI-generated by simply asking, ‘Was this generated using Google AI?’
Gemini employs a scan of the video’s visuals and audio to detect Google’s unique watermark known as SynthID. This functionality, previously introduced for images in November and restricted to content produced or modified with Google AI, serves as a distinctive verification method. Gemini can pinpoint the exact moments in the video or audio where the watermark is present, going beyond a simple yes or no response.
While some watermarks can be easily removed, Google asserts that its SynthID watermark is nearly imperceptible. The effectiveness of removing this watermark or its detectability by other platforms remains uncertain. Notably, Google’s Nano Banano AI image generation model within Gemini embeds C2PA metadata, but the inconsistent tagging of AI-generated content on social media platforms enables the circulation of undetected deepfakes.
Capable of verifying videos up to 100 MB and 90 seconds in duration, Gemini’s enhanced feature is accessible across all languages and regions where the app is available.
Source: The Verge
Leave a Reply