Spotting AI's work can be increasingly difficult as its capabilities and subtleties continue to improve. This continued shift makes labeling AI generated work all the more critical — something that is being done in bits and pieces. The latest development to do so comes from Google, which will now use SynthID technology to mark mages edited using Reimagine in Magic Editor.
Google DeepMind launched SynthID in 2023, a technology that allows for imperceptible digital watermarks within any content created with generative AI. The company has previously used it in AI-powered programs such as Lyria, Imagen and Gemini.
Now, SynthID will be making its literal mark on Magic Editor. The tool allows users to reimagine their photos by moving, erasing or adding special effects to different aspects of the image. The Reimagine feature is available for anyone 18 or older and on the Pixel 9 or later. Users who click Reimagine can enter a prompt to change bits of the image. Google recommends entering specific, simple phrases, such as "autumn leaves," and then explore potential options before picking one. The only difference is now that image will have a watermark from SynthID. It's available for anyone to see by clicking "About this image."
This article originally appeared on Engadget at https://www.engadget.com/ai/googles-magic-editor-will-watermark-its-ai-tweaked-photos-170009281.html?src=rss https://www.engadget.com/ai/googles-magic-editor-will-watermark-its-ai-tweaked-photos-170009281.html?src=rssConnectez-vous pour ajouter un commentaire
Autres messages de ce groupe


In case you need further evidence that companies involved in AI chip manufacturing are looking to give themselves whatever edge they can, SoftBank is adding to its institutional expertise by buying

With the Amazon

Apple announced a bunch of news recently, as the company


Any music fan will know how time-consuming it can be to keep tabs on which artists are playing nearby soon. To make that a bit easier — and help make sure you don't miss shows by musicians you like
