Photo by Mister Mister on Pexels

Google has introduced SynthID Detector, a tool aimed at bolstering content provenance by identifying material created using its AI models like Gemini, Imagen, Lyria, and Veo. Revealed at Google I/O, the detector analyzes various media types including images, audio, video, and text for subtle SynthID watermarks embedded by Google’s AI generation tools.

The SynthID Detector portal highlights segments within content that exhibit the strongest likelihood of bearing these digital watermarks. For audio, it precisely identifies watermarked sections, while for images, it pinpoints areas where watermark detection probability is highest. This granular analysis allows users to more efficiently assess the authenticity of the media.

Currently undergoing initial testing, Google is gradually expanding access to SynthID Detector via a waitlist. The company emphasizes that the phased rollout will enable them to collect valuable user feedback and optimize the tool’s performance before wider deployment. The effectiveness of the tool in real-world scenarios is still under evaluation.