Yeah, I’d say the version you got has been upscaled and had the canvas expanded by AI. I see, for instance, that the can’s size labeling has been changed to the American version, the Hangul below the logo looks wrong, and the “Hello” goes slop at exactly the point where the original picture cuts off.
Some interfaces use fixed aspect ratios, so perhaps the diffusion model automatically outpainted the original image from the bottom to make the edit, even if that wasn’t the intent.
Yeah, I’d say the version you got has been upscaled and had the canvas expanded by AI. I see, for instance, that the can’s size labeling has been changed to the American version, the Hangul below the logo looks wrong, and the “Hello” goes slop at exactly the point where the original picture cuts off.
What a strange choice to burn tokens on.
Some interfaces use fixed aspect ratios, so perhaps the diffusion model automatically outpainted the original image from the bottom to make the edit, even if that wasn’t the intent.
OP’s image uses the correct aspect ratio though.