3 ms·

These are beautiful. Yet... I know this sounds arrogant, but I feel like I would eventually realize they're real photos if I stared at these long enough withou

These are beautiful. Yet... I know this sounds arrogant, but I feel like I would eventually realize they're real photos if I stared at these long enough without being told. i.e. If I saw them framed on a wall and walked past them every day. The image noise, surface imperfections, chromatic aberrations, etc. are clues, and yes a raytracer can reproduce that, but not usually this well. That's precisely why I think these are beautiful.

viraptor 4y ago
I think the main thing that gives away the shapes image is that the glass surface has thickness. You can see that there are 2 layers of reflection visible on horizontal edges because of it. You could do that in CGI, but nobody does - they make the base surface slightly reflective. I don't think I would notice it out of context though...
- startupsfail 4y ago
  RTX does it in real time, in some settings. And then DLSS can infer on top.
- berkut 4y ago
  In VFX at the high-end (the smaller studios do often take shortcuts) we do model full layered materials with thickness, at least for mid to hero-level assets, so we would do the glass with proper volumetric thickness, with fully-correct inter-reflecting/refracting dielectric BSDFs modelling the glass, with correct IOR and total internal reflection. It is slower to render than using "thin glass", and lighting the checkerboard surface underneath the glass isn't as efficient (you can't do naive Next Event Estimation through the refraction of the glass volume, but you can do manifold walks to solve that), but we do it at that level of fidelity often enough when required and have been for years. There are times when it's not required though, so approximations can be acceptable in more forgiving situations.
  - viraptor 4y ago
    Yes, you totally can do it if you want to. What I meant is that that kind of CGI scene composition does not include a glass sheet on the surface. They used it to add the reflection to a checkered pattern instead of trying to find something like polished marble. But if we found a thousand different versions of that scene in CGI, I would bet that every single one of them would use a checkered pattern plane that's slightly reflective. Maybe even uses marble subsurface scattering. But none would add a sheet of glass on the plane as a not-forced modelling choice. That's what makes the photo identifiable as a photo.
    - berkut 4y ago
      Sorry, I mis-understood: I agree.

viraptor 4y ago

I think the main thing that gives away the shapes image is that the glass surface has thickness. You can see that there are 2 layers of reflection visible on horizontal edges because of it. You could do that in CGI, but nobody does - they make the base surface slightly reflective. I don't think I would notice it out of context though...

startupsfail 4y ago
RTX does it in real time, in some settings. And then DLSS can infer on top.
berkut 4y ago
In VFX at the high-end (the smaller studios do often take shortcuts) we do model full layered materials with thickness, at least for mid to hero-level assets, so we would do the glass with proper volumetric thickness, with fully-correct inter-reflecting/refracting dielectric BSDFs modelling the glass, with correct IOR and total internal reflection. It is slower to render than using "thin glass", and lighting the checkerboard surface underneath the glass isn't as efficient (you can't do naive Next Event Estimation through the refraction of the glass volume, but you can do manifold walks to solve that), but we do it at that level of fidelity often enough when required and have been for years. There are times when it's not required though, so approximations can be acceptable in more forgiving situations.
- viraptor 4y ago
  Yes, you totally can do it if you want to. What I meant is that that kind of CGI scene composition does not include a glass sheet on the surface. They used it to add the reflection to a checkered pattern instead of trying to find something like polished marble. But if we found a thousand different versions of that scene in CGI, I would bet that every single one of them would use a checkered pattern plane that's slightly reflective. Maybe even uses marble subsurface scattering. But none would add a sheet of glass on the plane as a not-forced modelling choice. That's what makes the photo identifiable as a photo.
  - berkut 4y ago
    Sorry, I mis-understood: I agree.