Google introduces a new verification portal, SynthID Detector, to identify AI-generated text, audio, video, and images.
SynthID Detector provides users with a central place to analyze AI-generated content. Those who upload content created with Google’s AI models, such as Gemini, Imagen, Lyria, or Veo, will immediately see if that content contains a SynthID watermark. The system marks specific parts of the text, audio, or images where the watermark is likely present.
The new tool is currently available to a select group of testers, including journalists, media professionals, and researchers. They can join a waiting list to gain access. A broader rollout will follow later.
Invisible Watermarks
SynthID works with invisible watermarks that, according to Google, remain detectable even after edits or transformations. Initially, the technology was only applied to AI images, but the system has since been expanded to include text, audio, and video content. Google reports that more than 10 billion items have already been provided with a SynthID watermark.
Open source
To encourage the use of the technology, Google has also made the watermarking mechanism for text open source. This allows developers to integrate the system into their own AI models. Additionally, there is a collaboration with NVIDIA. Videos generated with NVIDIA “’s” Cosmos preview will now be automatically provided with a SynthID watermark.
Furthermore, Google announces a partnership with GetReal Security, a content verification platform. This collaboration aims to ensure that external parties can also recognize SynthID watermarks.