What underlying AI technologies and machine learning models typically power photo to comic conversion?

The primary underlying AI technologies are deep learning models, particularly Generative Adversarial Networks (GANs) and, more recently, diffusion models. GANs consist of a generator network that creates the comic image and a discriminator network that tries to distinguish it from real comic art, iteratively refining the generator's output. Diffusion models work by iteratively denoising a random noise image, guided by the input photo, to gradually produce the desired comic style. Other techniques like neural style transfer, convolutional neural networks (CNNs), and image-to-image translation models also play significant roles.