What you ran into is actually a pretty common quirk with a lot of generative image models, including tools like undress ai they don’t understand human anatomy the way we do, they just stitch together patterns they’ve seen in training data. When you see three belly buttons or a leg that looks like a baguette, that’s the model mixing up features because it’s predicting pixels, not reasoning about structure. These systems are impressive in some ways, but they still hallucinate — especially when the prompt or use case pushes them outside the kinds of images they were trained on. So results that look glitchy or surreal (like something out of a blocky game) are just the current limits of the underlying AI, not a sign of deep “understanding.” Other users report similar bizarre outputs when the model gets confused about context or anatomy.