Google upgrades Gemini AI with video verification using SynthID watermark, allowing users to see if content is AI-generated
Google's Gemini app now flags video segments generated with Google AI
Google has added a new video authentication feature to its Gemini app. Users can now see if a video was created or edited using Google's AI tools. This update is available globally for Gemini users. The move comes as governments, platforms and media organizations step up efforts to combat the proliferation of manipulated and synthetic video content online.
How the verification function works
This feature allows users to upload a video file to the Gemini app and ask if it contains AI-generated material. Gemini performs checks for SynthID, an invisible digital watermark embedded in content generated by Google's AI systems. So instead of getting a simple confirmation, the user gets a highlight of the video segment that includes the SynthID marker. This provides insight into which segments may have been generated or modified using Google's tools.
Also read: Disney confirms $1 billion investment in OpenAI: Gives Sora access to flagship characters
Range and availability
Google says the verification tool supports videos up to 90 seconds and files up to 100 megabytes. This feature works in all languages supported by the Gemini app and is being rolled out to users worldwide.
SynthID was originally developed to verify images and audio generated by AI. This announcement is another step in Google's efforts to make AI-generated content easier to identify.
Also read: India emerges as world's third most competitive AI powerhouse: Stanford
Why did Google choose to introduce this now?
Generative AI is under scrutiny for its role in creating misleading and fabricated media. Big tech companies are under pressure from regulators and policymakers to provide barometers for identifying AI-generated content. This particularly concerns video manipulation tools, which are becoming more sophisticated. Google says Gemini's verification feature is intended to provide users with more information about the origin of digital content, rather than as a final determination of authenticity.
