Celine IA photos

[CélineI'mAlive]

Hello everyone, guys. As some of you know, I’m a big collector of everything Céline does. I have hundreds of interviews and videos of her. I’m currently working on the photo section, but I’ve run into a problem: as Thomas shows, AI is creating increasingly realistic photos of Céline, which makes it harder and harder to know whether the photo I’m seeing of Céline on any website is real or fake. I know there are many collectors of Céline photos here. What would you advise me? Thanks.

[Thomas]

It would be a bit time consuming if you had thousands, but if you were suspicious of a particular photo you can, ironically, ask an AI to check it. If it was created by a Google product such as Nano Banana then they insert an invisible watermark called SynthID and this would be read during the check. It's not 100% fool proof as I have tested a few AI photos and if you edit them a bit after they were created, it can sometimes lose the watermark and it thinks it is a real photo.

 

But it still checks for inconsistencies on any image you upload even without a SynthID, such as weird text or something that looks wrong, or if the appearance of the person makes sense for the time period it claims to be in (such as a child Celine appearing with Ariana Grande, it wouldn't be logical). It also checks back to see how long the photo has been around for and if it has searched it in the past, stuff like that.

 

I would use something like gemini.google.com. As time goes on I think the transparency over AI will be addressed with unremovable watermarks and platforms like Instagram automatically identifying an AI image and adding an icon or something.

[CélineI'mAlive]

It would be a bit time consuming if you had thousands, but if you were suspicious of a particular photo you can, ironically, ask an AI to check it. If it was created by a Google product such as Nano Banana then they insert an invisible watermark called SynthID and this would be read during the check. It's not 100% fool proof as I have tested a few AI photos and if you edit them a bit after they were created, it can sometimes lose the watermark and it thinks it is a real photo.

 

But it still checks for inconsistencies on any image you upload even without a SynthID, such as weird text or something that looks wrong, or if the appearance of the person makes sense for the time period it claims to be in (such as a child Celine appearing with Ariana Grande, it wouldn't be logical). It also checks back to see how long the photo has been around for and if it has searched it in the past, stuff like that.

 

I would use something like gemini.google.com. As time goes on I think the transparency over AI will be addressed with unremovable watermarks and platforms like Instagram automatically identifying an AI image and adding an icon or something.

The problem is this: I have thousands and thousands of photos of Céline. I’ll do what you suggest, but randomly, like factories test products to check them. Thank you very much!