AI sucks at legible text in general. It can certainly do it, but very often the text very clearly does not belong there. You can usually tell it wasn’t written or printed there by anyone, it just does not fit very well. This image has very natural looking text in the lower right corner, on that bag thing.
AI sucks at consistent reflections in the eye. This dog has a pair of quite consistent reflections in its eyes. You can even see the phone that’s taking the photo in its right (our left) eye.
AI images are generated from noise. Some of that noise is leftover at the end of the process, especially when you’re rendering intricate stuff, like fur. That noise looks blurry, but it looks nothing like blur. I can’t see any of that in this image, it all the natural smartphone post-processing and compression feel, mixed with regular ol’ blur of being out of focus.
AI sucks at consistently aligning lines in the background when they’re interrupted by an object in the foreground. Like edges where the wall meets the floor, or edges where walls meet each other. Very often the AI does create a continuous wall behind the object, but the lines will be misaligned. None of that here.
AI is also not the best at drawing thin hair-like things that stick out against “open” background, like individual strands of hair or, in this case, whiskers. They often look wrong somehow.
Now none of those things is sufficient evidence by itself because AI can and will occasionally get those right, but together all at once, they make a convincing case that this image is altered at best, but probably not AI generated.
I don’t think it is, honestly.
AI sucks at legible text in general. It can certainly do it, but very often the text very clearly does not belong there. You can usually tell it wasn’t written or printed there by anyone, it just does not fit very well. This image has very natural looking text in the lower right corner, on that bag thing.
AI sucks at consistent reflections in the eye. This dog has a pair of quite consistent reflections in its eyes. You can even see the phone that’s taking the photo in its right (our left) eye.
AI images are generated from noise. Some of that noise is leftover at the end of the process, especially when you’re rendering intricate stuff, like fur. That noise looks blurry, but it looks nothing like blur. I can’t see any of that in this image, it all the natural smartphone post-processing and compression feel, mixed with regular ol’ blur of being out of focus.
AI sucks at consistently aligning lines in the background when they’re interrupted by an object in the foreground. Like edges where the wall meets the floor, or edges where walls meet each other. Very often the AI does create a continuous wall behind the object, but the lines will be misaligned. None of that here.
AI is also not the best at drawing thin hair-like things that stick out against “open” background, like individual strands of hair or, in this case, whiskers. They often look wrong somehow.
Now none of those things is sufficient evidence by itself because AI can and will occasionally get those right, but together all at once, they make a convincing case that this image is altered at best, but probably not AI generated.