It would be nice to have a Megascans system at home and it seems that idea is not so far fetched.
This research paper comes with the full source coude:
https://mediatech.aalto.fi/publications/graphics/FourierSVBRDF/
Video presentation here (time 27:36):
Unfortunately most code is in Matlab (universities, sigh) so in order for it to be practical it has to be ported to c++.
Anyone interested?
Or alternative someone with a license for the MATLAB Coder (a tool that converts matlab to C) could convert it.
Yeah, I looked into that but it doesn’t seem as straightforward as you would hope.
PBR scanner… lol… this is most like an brute approximation.
PBR in general is far far away from real model. I wonder if those guys around the world making “PBR scanners” have any physics knowledge (I am not referring at that link, in general). The light real model is much more complex.
It doesn’t have to be physically correct if it’s close enough. And for games, “close enough” often just means “looks cool”.
I mean, I don’t know about your team, but over here my artists certainly haven’t studies physics in order to generate the textures we’re currently using. And no decision about what to do has ever come down to “so what’s the closer approximation to how light really works?”
So if the scanner can get something that a) looks decent and b) is tunable to fit into our (not realistic anyway) art direction… sweet.
Well, from simple diffuse to PBR is a big jump. Close to real model? not really. Maybe in the future when a video card can do 1000x times what do today we can talk about something close to real model.
So… 2 generations after the nVidia pascal? 
At the rate memory has been increasing in graphics cards, I’m a bit scared to see what we’ll have in two generations. 
Well that’s good for dreamers but most people will make do with PBR which is the fastest compromise for realistic lighting we know of today.
Weyl fermions!!
@Kiwasi any new understanding on this?
What? You mean it’s not 100% physically accurate? Unity doesn’t construct the environment and lighting out of trillions^trillions of particles!? Baked lighting isn’t made by letting virtual photons bounce around the scene a million times? Objects don’t rotate because of spinons and orbitons? 
Assuming a generation means every 6 months or a year, In 2 generations we should see a 20% increase over what there is today. GPU and CPU manufacturers have several buckets in which they throw products, ie consumer, enthusiast, professional, server etc…
And every “generation” we see at absolute maximum a 15% hike in performance. It’s usually closer to 10%. This is not moore’s law specifically as there’s much faster available, just much more expensive for them and you. The 10% model works really well financially.
You can see this with benchmarks on all sites, whenever a new GPU or CPU is released in the same category ie consumer - the hike over the competition of last gen is usually around 10% more performance.
I don’t think this is a coincidence after watching it for years, and given the fact they can and do have much better hardware available in bespoke fashion, one can assume it’s market dynamics at work.
Actual processing performance may only increase marginally, but graphics memory appears to be increasing in leaps and bounds. Current reports indicate that NVIDIA’s Pascal is moving to HBM2 with a maximum capacity of 16GB so it wouldn’t surprise me if the x80 and x70 cards went up to as much as 8GB.
Yeah it’s really required for 4k although actually running it well at those resolutions is still not a guaranteed thing. I am developing a PS4 game. This thing has an awesome amount of available memory. It supports 4k displays, but not for gaming.
I’m not complaining, I can always find good reasons to burn more ram
Haha nice irony!
Basically, when we will have the power to compute quatrazillions electromagnetic waves per atom reflection/refraction, for quatrazilions^quatrazilions of atoms representing a small face of a rock, with density elements accordingly, we can say we are close to real model. Until then we just lie our selfs that PBR looks “realistic”.
Sorry guys. I’ve kind of dropped the ball on this one. Lately my understanding efforts have been going into more mundane problems. Specifically how can we provide enough food for the 7 billion odd people on this planet.
My part of that question has recently amounted to how do I stop chlorine production from shutting down every three days. Which in turn becomes how do I convince IT that upgrading to 64 bit is totally needed for any sensible manipulation of 1 GB data sets.
I will take your just criticism of my priorities on board and amend my ways. 
I sure hope so, the future of technology depends on you! probably.