Wow, a few years ago, if you've shown me this and Genie 3, I'd assume there were at least 10 years of development between them. This looks worse than Doom.
I'm pretty sure that got debunked as a handful of overfitted examples that fall apart as soon as you leave the area that was 3D recorded. Google's new tech is on another level entirely.
Hey, this looks really cool! I'm wondering - what happens if you feed it something totally different like a Van Gogh painting or anime character? Have you tested any non-photo inputs?