Once trained, the model can “predict” what the body might look like underneath clothing by replacing the clothing area with simulated skin. For more on the specifics of how it works, check out https://saucesenpai.com/blog/explained-what-are-deep-nudes/. GANs work well for this because they have two networks that improve each other—one generates the altered image while the other critiques it for realism, refining the result. This process requires significant data, and although the results are technically impressive, it raises serious ethical concerns. 

I creates deep nudes through a type of deep learning model, often a generative adversarial network (GAN) or similar neural network architecture. These models are trained on large image datasets, learning to recognize body shapes, skin tones, textures, and lighting patterns.