I've seen a lot of terrible analysis of this photo.
People are either shoving it through an AI detector, asking Grok, or they are peeping pixels with expertise they don't have.
Like, you can't go drawing straight lines on something with a highly skewed perspective that also has a distorted shape.
Or upload it to a sketchy AI detection website!
They often have the accuracy of a coin flip. Even lab-grade tools are only 70 to 80% accurate in controlled conditions. And yet people are trusting a free website with an ad for boner pills in the corner to tell them if something is authentic.
I have been doing high level photo manipulation for two decades. I'm as close to an expert as you will get on Tumblr dot com.
So let's properly peep at the pixels and do actual forensic analysis.
First, I think this is mostly a real photo. Probably taken at some other point in time.
And I don't think this is fully AI generated. I actually think it is a traditional composite. I think the hand and newspaper are separate assets that were blended. It's possible the hand was AI-generated and then composited. And I think they may have taken a real photo of someone holding a newspaper and replaced the hand.
The first oddity is the fingers.
Typically when you touch an object it creates a contact shadow. One finger has a contact shadow and the other does not.
It should probably look more like this.
The next sign of a composite is the edge of the finger.
There is a sign of a feathered edge.
This is a lazy compositing technique to help edges blend without making a super precise selection. If you look at all of the other edges in the photo, this is the only one that has a feathered edge. You can see how clean all the other edges are by the top arrow and how fuzzy the finger edge is by the bottom arrow.
If I were cutting out his other hand and taking the time to do it properly, I would clean up the edge to make sure it was consistent with everything else in the photo.
If I were in a hurry, I would just feather the edge and hope no one actually zooms in.
And then there is the edge of the newspaper.
This is called a matte line. The newspaper was most likely against a dark background when they cut it out, and they did not clean up the edge.
Again, it's lazy. Because Photoshop has a tool dedicated to fixing this exact issue.
In my expert opinion, I think they generated an AI hand, took a photo of someone holding a newspaper in similar lighting, and then manually blended them into an existing photo.
But I don't think this was 100% AI-generated. You typically don't see compositing errors in generated images. They probably couldn't get the AI to generate the newspaper without garbled text.
What's curious is that the pixel resolution is just barely bad enough that you cannot tell if the text is authentic. But it's not blurred or distorted. It is just low enough in resolution to give a sense of text without being legible. And I think that made people suspicious due to AI's reputation when text is involved.
But from what I can tell, the print size and letter spacing does seem to match.
So I don't think that is the clue people are making it out to be.
Last thing, image analysis like this is not 100% conclusive. I'm pretty sure there are shenanigans, but anyone who tells you with absolute confidence that an image is fake... is probably bullshitting or ignorant.
The missing contact shadow could be explained by the angle of the light filling it in.
The feathered edge could be motion blur.
The edge of the newspaper could be a sharpening artifact.
But the fact that the hand and the newspaper were vital aspects of the photo for proof of life, those three variables make this really damned suspicious.
discoholicmusic
s-u-f-j-a-n-s-t-e-v-e-n-s
pixelskylines
romaincrisis
origamiaround
