Google originally launched SynthID, its digital watermark for AI-generated content, as a way to detect whether an image was created using the company’s Imagen model in 2023. Now, at Google I/O 2025, the company is introducing a public-facing tool called SynthID Detector that claims to detect those watermarks in just about anything you upload.
SynthID Detector will be available as a web portal where you can upload images, video, audio and text to be scanned. Once uploaded, SynthID Detector Google claims the portal can tell you whether your upload contains AI-generated material and even “highlight specific portions of the content most likely to be watermarked.” For audio, the tool is supposed to be able to identify a specific portion of a track that contains the watermark, too.
SynthID was designed to mark content from Google’s models but Google hopes other companies will adopt the watermark for their own AI output. An open source version of SynthID is already available for text watermarking, and as part of the rollout of SynthID Detector, Google is partnering with NVIDIA to mark media its NVIDIA Cosmos model generates. SynthID Detector won’t be the only tool that can spot Google’s watermark, either. The company says GetReal Security will also be able to verify if media contains SynthID.
Considering the sheer number of ways Google hopes people will using AI to create images, video, text and audio, from the Audio Overviews in NotebookLM to short films made with its new Flow tool, it makes sense that it would offer a way to know if any of those things are real. Until models from one company produces the vast majority of content or a digital watermark reaches widespread adoption, though, a tool like SynthID Detector can only be so useful.
Journalists, researchers and developers can join a waitlist to try SynthID Detector through Google’s online form.
Credit: Source link