VrayMtl: Don't affect Alphachannel checkbox

[Dschaga]

It's a rendering not RW example. ..anyway you are right with the internal reflections of the edges ..they are to strong.

you mean fresnel?


I'm pretty sure that the reflections don't influence the fresnel of the density which is what we see in the alpha channel. If this is wrong, then a reflection of a very bright object would influence the density?

What the refractions do is a differnt kind of question, but my opinion is, that also refractions don't influence the density, because refraction ..or the look of the refraction... is for itself only related to the density (thickness and form) of the glassobject. ..agian if it would be wrong ,this would mean the alphachannel of a sphere on a bright floor and dark background would be inconsistent.

[studioDIM]

I guess the problem is more about editability in post, than correctness.

Having a white alpha due to refractions might mean more paintwork is some specific cases.

Then again, it's a problem the studio i work for faced just a couple of weeks ago (we had glass "sails" on a beach, with their edge visible), and we solved it as i described, with a couple more passes (one full on, the other simple fresnel), extremely quick to render, despite the massive size of the images (4k x 1600+).

Should one use other file formats than straight RGBA, though, there would be little or no need for additional passes (see rpf + combustion, for instance).

[Dschaga]

I guess the problem is more about editability in post, than correctness.

well, ...it started because editability, but becomes now really theoreticaly.
The alphachannel is a 100% CG problem and will not have any similarity in RL .. so the question is what the alphachannel will represent.

btw, i remember a renderfx which replaces the alphavalues of defined objID's.

[vlado]

The alphachannel is a 100% CG problem and will not have any similarity in RL .. so the question is what the alphachannel will represent.

If this is well defined, it should be easy enough to implement it. However, I'm still puzzled as to what exactly you want in the alpha channel, so any pointers and examples there would be very useful.

Best regards,
Vlado

[Da_elf]

your not the only one still confused

[dlparisi]

Looking back at Dschaga's image post, it looks like it is the refraction that is causing the change in the alpha, not the reflection. This seems perfectly correct, in RL and CG. The ray traced through the glass does not exit directly through one plane of glass and out the other side, it's actually bending and passing through several layers of the glass, so the alpha would be building up. The bend effect is just exaggerated at the corners because it is passing through so much glass (down the length of it) so it bends through a completely different part of the model.

Does the problem go away if you turn the IOR to 1?

[Dschaga]

mmh!
Dlparsi, you are right!
IOR can influence the density/alpha optically too - i'm only unsure if vray maybe mixup refraction and reflections because i always thought fresnel will only change the behave of the reflections and when you disable fresnel reflection you get a completly different alpha.


Vlado, it's really difficult to come with RL examples, i only know that vray handles Glass/alphachannel very different to lightscape, but lightscape was always a reference when it comes to CG glass and how to handle this output in PS. ..and to be honest ..this thread should be in the bug forum and not wishlist.

Anyway, something which gave me a different view on glasstransparency since there is also some kind of psychological effect - was the use of a Circular polarization filter:
It's strange but a bright object which is a reflection in a window will make a window less transparent and a dark Background will don't even influence the transparency in RL. (remember i was always talking about reflections shouldn't effect density ...err, was i wrong now?)

Now with this polfilter you can completly fade out the reflections and look straight through the window into the room just if there wouldn't be any glass. ..this is crazy, but it's logicaly because a Circular Polfilter can prevent rays from the reflections to hit the film on the camera.

When i keep this in mind, then reflections should be influence the alphachannel but only dependent on the grade of brigthness. .. duh ..because this will make PS postprocessing really complicated.

I also recognised on some specific situations, that the reflections tend do loose saturation, but this is maybe related to the used fassade.

I hope i can shoot some examples tomorrow with a digicam, because i only have used it on my f3.

[studioDIM]

well, ...it started because editability, but becomes now really theoreticaly.
The alphachannel is a 100% CG problem and will not have any similarity in RL .. so the question is what the alphachannel will represent.

If Alpha in CG = ability to see THROUGH in RL, then Vray does it right (when the refraction depth is set high enough).
Besides, MentalRay does it in exactly the same way (with the Glass Phenomenon.), some noise aside.

I wouldn't change the engine, if i can generate one more pass (fresnel map on opacity to get the LS type of Alpha, MtlWrapper to get scanline type of Alpha, RPF with Obj, MTL, or rendernode ID to select the object with a click in post) and get precisely what i need.

btw, i remember a renderfx which replaces the alphavalues of defined objID's.

It's probably much simpler to output the ObjID pass and do it by hand in post.
It surely is more flexible.

[Dschaga]

here the example.. it's could outside and it snows so i decided finding a motive at home... well , here we are:

Circular polfilter @0°:


Circular polfilter @90°:


The question now, would be : is the transparency different or does it only looks like different because the reflections are hidden on one image?

[studioDIM]

Lol, i guess in the first pic you see actually LESS of the wall behind the window.
That alpha would be grey.

In the second, polar bears (heum) or not, the glass is entirely transparent.

I can't see what's the problem