Am 21.06.2015 um 09:44 schrieb Nekar Xenos:
> Can some explain to me what exactly PBR is. In Pov-Ray terms would be
> nice :)

PBR avoids old-school shading models and rendering algorithms that were 
designed to achieve a particular effect, and instead uses mechanisms 
that were designed to model the underlying physical processes.

POV-Ray has been heavily geared towards PBR in recent times, and UberPOV 
should by now be a viable PBR tool, provided you stick to the following 
rules:

- Don't use "ambient". Use radiosity instead, and "emission" where 
applicable.

- Don't use "phong" or "specular". Rely on reflections instead (which, 
for good results, mandates that you use blurred reflections).

- If you decide that you do need to use "phong" or "specular", prefer 
"specular", and /always/ use it in tandem with reflection, making sure 
that the parameters match: Use "specular albedo N" where N is the 
maximum reflection intensity, and use "fresnel on" or "metallic on" for 
the entire finish block just as you do in the reflection block. Also, 
make sure to specify the same "roughness" in the reflection block. 
(You'll need an up-to-date version of UberPOV for "fresnel on" in the 
entire finish block and "roughness" in the reflection block.)

- Always use "fresnel on" for reflections (don't forget to specify 
"ior"), or "metallic on" where applicable, and always use 0.0 for the 
minimum reflection (or leave it out, specifying the maximum reflection 
only).

- Don't use colours as reflection parameters. Use the "metallic" 
mechanism for this instead.

- Don't use "light_source". Use objects with "emission" (or a sky 
sphere) and radiosity instead.

- If you decide that you do need to use "light_source", always use a 
well-matching "looks_like" object with "no_radiosity on", always use 
"area_light", always use "fade_power 2" with "fade_distance" set to 
(roughly) half the area light vectors' dimensions, and always use 
"area_illumination" unless the light source is sufficiently far away. 
Also, do make use of photon mapping.

- Use "normals on" and "media on" in the radiosity settings. Use 
high-quality settings, and/or turn sample caching off (requires UberPOV).

- Don't use "filter" (unless you know what you're doing). Use a fading 
interior instead, with "fade_power 1001", or use absorbing media.

- Don't use "rainbow" (does anyone?).

- Do use a little bit of reflection on all your materials (typically 
with heavy blur).

- If something doesn't look right, don't ask "what effect can I add or 
tweak to make it look better?" but "what phyiscal properties of my scene 
have I not modelled correctly?"


As for HDR light probes, contrary to what Jerome wrote, the question of 
whether they're legit as a /sky sphere/ to avoid complex modelling is 
outside the scope of PBR; what's a no-go in PBR is using them for 
/environment mapping/ to fake reflections, but POV-Ray doesn't support 
that technique out of the box anyway. (You /could/ achieve a similar 
effect and possibly some of the speedup by using a HDR sky sphere and 
setting all objects to "no_reflection on", but environment mapping is 
also frequently used with different maps for different objects.)


The advantages of PBR are threefold:

- It takes far less time to achieve a convincing result, provided you 
have a basic understanding of the the underlying physical effects, 
because you don't need endless iterations of tweaking your material and 
light source parameters, repeatedly trying to figure out (A) what's 
still wrong about the resulting image, and (B) how to counter this with 
the knobs available (without introducing other undesired effects). With 
PBR, first of all your images will /always/ be physically consistent in 
themselves (and thus convincing in a very basic sense), and second, all 
you need to achieve a truly convincing look is to carefully examine the 
materials and light sources you intend to model and plug in the 
corresponding values. If you did it right, all you have left to do is 
worry about whether the image is pleasing, not whether it is convincing.

- You can freely combine elements from different sources. In a non-PBR 
environment, only the combination of materials and lighting can be 
assessed for whether they are convincing or not, and any attempt to set 
up some "neutral" setting to assess just one or the other is moot: To 
assess whether a lighting setup is "neutral" it would have to be tested 
with known "neutral" materials, but to assess whether a material is 
"neutral" in the first place it would have to be tested with a known 
"neutral" lighting setup - a classic hen and egg problem. In practice, 
different authors will inevitably have different ideas what a chicken 
actually is. In a PBR environment, however, a material can be assessed 
simply by checking whether the parameters plugged in match the optical 
characteristics observed in reality, without rendering even a single 
image, and the same goes for lighting setups. So even if you don't have 
the facilities to measure the exact optical chracteristics of all your 
materials and lighting setups, you can jump-start your chicken farm from 
a small set of precisely known materials and lighting setups.

- You can rely on the results being physically accurate, allowing you to 
model scenes for which you have no real-life reference image to go by. 
(Maybe one of the most spectacular examples is the black hole and 
accretion disk in the "Interstellar" movie, which was modeled solely 
from physical principles - at least as far as the distortion of 
spacetime is concerned - rather than preconceived notions of how such a 
phenomenon would look like.)

Post a reply to this message