I’m looking for them to name some tricks to reverse engineer, they are tight lips except something about automated decimation … I wonder if this will scale to any 3d games and where are teh real constraints. I have some idea, it might be an advanced version of the box environment mapping with spatial positioning, but where the box is replace by a low poly view based approximation of the scene to conserve parallax data (dependent of the navigation volume) …
Haven’t watched the video, but I have played around with the new Elemental demo for daydream which has a new renderer in it which I assume is this? If it is, the performance does seem to be pretty decent on my Moto Z. Obviously they are demos so it might not hold up to real world stuff, but even to have the demo with the lighting was pretty good for mobile VR.