Photo credit: www.androidcentral.com
What you need to know
Google has introduced an important update to its “Reimagine” feature in Google Photos, which enables users to modify their images using generative AI technology. This month, the company is implementing SynthID, a digital watermarking system, to denote images generated or altered using Reimagine. Users can verify if an image was created entirely with AI or modified through this tool by utilizing Google Lens or Circle to Search.
Last year, when Google launched the AI-driven “Reimagine” tool alongside its Pixel 9 series, it sparked significant discussion. While it allowed users to create imaginative and sometimes misleading alterations of their photos, it raised concerns about the authenticity of images, leading to public scrutiny. In response, Google later announced plans to watermark AI-enhanced photos in Google Photos.
The company has stated, “Beginning this week, Google Photos will start utilizing SynthID, a technology designed to embed an imperceptible digital watermark into generative AI content within Reimagine. This facilitates quick and easy identification of AI-generated content.”
SynthID has been integrated into various Google AI applications as a means to effectively detect and label AI-generated materials. By embedding watermark data directly into the content, the technology remains undetectable to the human eye yet provides a straightforward identification process. This approach is intended to help differentiate AI-created images, music, and text from human-generated content.
For instance, images produced by Google’s Imagen 3 text-to-image generator already feature SynthID digital watermarks. However, images only partially modified using Reimagine hadn’t previously been subjected to this watermarking process, making it difficult for users to discern genuine images from AI-manipulated ones. Now, when users question whether an image has undergone AI editing, they can access the “About this image” feature in Google Lens or Circle to Search. This function will indicate any SynthID watermark details and provide relevant context, alongside the image’s metadata, which may reveal more about the modifications made.
Despite these advancements, not every image will be labeled. Google advises that “in certain scenarios, edits applied using Reimagine might be too minor for SynthID to detect—such as altering the hue of a small flower in the background.” Nonetheless, the implementation of SynthID for reimagined images is a substantial step forward in ensuring transparency and preventing AI-altered photos from being misrepresented as truthfully captured visuals. The new watermarking for images edited with Reimagine is set to take effect starting this week.
Source
www.androidcentral.com