Google has officially launched a new feature for the Gemini app, allowing users to directly check whether a video was created or edited by Google AI by simply uploading it. This update marks another major advancement for Google in the field of AI content transparency, helping users more reliably identify potential AI-generated media and address the growing challenge of deepfakes.

image.png

 Core Feature: Comprehensive Upgrade of SynthID Watermark Detection

Gemini app uses Google's proprietary SynthID digital watermarking technology (introduced in 2023 and embedded in over 20 billion AI-generated pieces of content) to scan uploaded videos. This watermark is imperceptible to human sight and hearing but can be accurately identified by specialized algorithms. Gemini will check both the video's visual track and audio track and provide detailed feedback rather than a simple "yes/no" conclusion.

For example, Gemini may return information like:

"SynthID watermark detected between 10-20 seconds in the audio. No SynthID watermark detected in the visual effects."

This detailed report allows users to clearly understand which parts of the video may involve AI-generated elements, greatly enhancing the practicality of content verification.

 Low Usage Barriers, Global Coverage

Users simply need to upload a video file in the Gemini app (supporting up to 100MB and 90 seconds in length), then ask questions like "Was this video generated using Google AI?" to trigger detection.

This feature is available in all languages and countries/regions covered by the Gemini app, without any additional subscription or tools required.

 Background and Significance: Promoting Responsible AI Development

SynthID is Google's long-term content tracking technology, now extended to images, audio, text, and video. The launch of this video detection feature follows the image verification function introduced in November, further improving Gemini's content authentication system. Google states that this helps enhance trust in digital media, especially in news, social media, and entertainment, enabling users to quickly distinguish AI-intervened content and avoid misinformation.

In the future, Google plans to further expand support for the industry standard C2PA content credentials, enabling detection of content generated by more non-Google AI tools.

AIbase Commentary: In an era of rampant AI-generated content, this free and convenient tool is undoubtedly a blessing for users and creators. It not only demonstrates Google's commitment to AI ethics but also sets a new benchmark for transparency in the industry. For more details, please follow updates on the official Google blog.