Google opens SynthID Detector to the public for AI media checks

Google has launched SynthID Detector, a public website for checking whether images, videos, and audio files contain its AI watermark. Ivan Mehta reports for TechCrunch that the service is now available to everyone after an earlier testing period for journalists, media professionals, and researchers.

The site accepts a broad range of commonly used files. Image support includes JPG, PNG, WebP, AVIF, HEIC, TIFF, and GIF. Users can also submit MP4, MOV, and WebM video files, plus audio formats such as WAV, MP3, OGG, FLAC, AAC, and M4A.

SynthID Detector looks for SynthID, Google’s watermarking technology. Google introduced SynthID in 2023 to mark media made with generative AI. Unlike visible labels, the watermark is intended to remain embedded in a file and can be checked through compatible verification tools.

Google applies the technology to output from its Nano Banana, Veo, and Lyria models. It also uses it in products with media generation features, including Gemini, Flow, ProducerAI, and Vids. For content teams, the detector can provide an additional signal when assessing material received from contributors, partners, or audiences.

Watermark checks have limits

Google says SynthID verification is already built into the Gemini app and Chrome, and that people make about one million verification requests each day. Other companies are also adopting the technology. According to TechCrunch, OpenAI, Nvidia, and Kakao support SynthID, while Apple is reportedly preparing support.

However, a detector can only identify watermarks that are present and recognizable. A negative result does not prove that a file was created by a human.