thanks! those are photographs of non-existing Santas generated with diffusion models like Dall-E, Midjourney or SD, animated with a GAN-like architecture. There is no depth involved as for other methods you mention, just user-rgb-in -> Santas out.

There is a simple voice changer as post-processing but nothing too fancy.