let's take a look inside

Yay!!!!

This is pretty impressive!

@zillowgonewild 👀

So this is near realtime interactive AI generated video right? Not gaussian splats?

Does it use the floor plan too, or just the listing photos?

just the photos, so far. we compute relative camera poses as a preprocessing step.

How much time does it take to generate this world?

this is a screenrecording of live interaction streaming from cloud GPUs. can do this after around ~30s of preprocessing to compute camera poses for the input images.

Impressive, is it open?