Really excited to share the beta release of #SynthID, a watermarking tool created by Google DeepMind and made available via Google Cloud to help tag and identify AI-generated images

One of the cool things about #SynthID is that it uses a new type of watermark embedded directly into the pixels of an image, which means the watermark is imperceptible to the human eye & detectable even if edited by procedures like cropping or filters

For now #SynthID will be available to a small group of Vertex AI customers using Imagen, @Google’s text-to-image model. This beta release will help us gather the feedback we need to continue improving the tool & bring watermarking to more Google products

While #SynthID isn’t a silver bullet for misinformation, it’s a promising technical solution to some of today’s pressing AI safety issues

@GoogleDeepMind @googlecloud Impressed by the new watermarking tool, but have you considered attacks like regeneration? Our research proves these can effectively remove any pixel-based invisible watermarks, regenerating images closely resembling the original. Paper:

I'm assuming that this won't be open-source? Given that it's mentioned in the article that that would make it easier for nefarious entities to get around this? I do wonder if there's a potential as well for general use where artists and creators can watermark digital art or products in a way that is imperceptible to the human eye?

@GoogleDeepMind @googlecloud 👏👏👏

@GoogleDeepMind @googlecloud Kudos, v.impressive. PS: Is it theoretically possible to do the same for audio?

@GoogleDeepMind @googlecloud What’s next for text?

@GoogleDeepMind @googlecloud I'm grateful for the tool. I'd like to be more excited about tagging generated content but the truth is I want to see better tools for creating generated content at this stage.

@GoogleDeepMind @googlecloud perfecto