They didn't say much about it but hopefully they can infer enough to do a good job. You can already do mildly ok fake depth maps from a single image. I also would like to know if this is going to be somehow locked down to Apple Vision or a format we could use in a Quest3 or other VR headset.