Terrible water quality in PT

Started by N-drju, June 13, 2020, 01:25:13 PM

Previous topic - Next topic

N-drju

I have a serious dilemma regarding the PT rendering. It seems that most of the assets in my project benefits from the PT rendering. Safe for water, which is just abominable...

I think that the volume density is the root of the problem here. While I can get a nice, wet, twinkling appearance of the water in a standard render, the PT version of it looks like a damn, dusty carpet. ::)

I have tried all of the available lighting methods for PT, in the water settings. Every single one of them results in something like this:

water.jpg

Let me say, this was not what I expected when I thought "PT" in relation to water. It looks like I am forced to decide whether I want to have deep shadows or a decent water... ::)

Sorry for the small sample size. I just don't want to overburden the system in the midst of a heatwave.
"This year - a factory of semiconductors. Next year - a factory of whole conductors!"

WAS

Well, the fact you're just switching between PT and standard is the issue. You also have to deal with lighting issues in shadows where noise will show. Higher quality settings.

PT water needs more customized density settings. Which is why the depth looks less intense in PT, because the default effect is lessened from stanadard.

I do notice the shadows look terrible still. There is no transmission, which to me is a bug and wrong, as PT should most certainly be offering this.

N-drju

To me, the fact that standard renderer is less noisy than PT in volumetric setting speaks volumes... Again, this is not how I imagined this. I wouldn't say that PT is "production ready".

What do you mean that switching between the options is a part of the issue? The settings are, in fact, customized - a density shader is added as a function to guide the "volumetric" effect. Anything else I could try to salvage this?
"This year - a factory of semiconductors. Next year - a factory of whole conductors!"

WAS

Quote from: N-drju on June 13, 2020, 02:25:46 PMTo me, the fact that standard renderer is less noisy than PT in volumetric setting speaks volumes... Again, this is not how I imagined this

You should learn about Path Tracing, noise is a inherent byproduct. That's why so many come with denoising algorithms on-top of AA. It's unfortunate. Like the current live renderers for PT are just noise shows.

Quote from: N-drju on June 13, 2020, 02:25:46 PMWhat do you mean that switching between the options is a part of the issue? The settings are, in fact, customized - a density shader is added as a function to guide the "volumetric" effect.

Yeah this is the issue I often have. A water that looks good in Standard, often doesn't look right in PT, like very dark. I have to fine tune the shader in PT, for PT.

I'm not exactly sure what youre settings are so I'm not sure what is really influencing this (lighting or shader settings).

Can you paste your water shader here? I can fiddle with it.

N-drju

#4
Quote from: WAS on June 13, 2020, 02:50:32 PMYou should learn about Path Tracing, noise is a inherent byproduct.

I know that of course - randomness sits at the core of path tracing. But the fact that you can reduce the noise in the cloud medium while being sucker-punched in the water medium is a bit strange. ??? Wouldn't you say?

Quote from: WAS on June 13, 2020, 02:50:32 PMCan you paste your water shader here? I can fiddle with it.
I'm afraid I can't because that would amount to compromising my contest entry, (might be considered as a teamwork then) so I'll pass. But thank you for the offer. Perhaps after I submit my work.
"This year - a factory of semiconductors. Next year - a factory of whole conductors!"

WAS

Quote from: N-drju on June 14, 2020, 02:09:28 AM(might be considered as a teamwork then) so I'll pass. But thank you for the offer. Perhaps after I submit my work.
I can guarantee you it won't. I've already provided help to other contestants regarding TG. Remember, you are allowed to use third party assets. Most of these are created by someone else. And I really just intend to provide you with suggestions to remedy the issue.
I think the noise issue is because clouds are being rendered differently than the environment. Similar to how you can change the render method of an object, I think clouds are handled differently.

N-drju

I don't know... I'll think about it. :-\

In the meantime, I have made several tests and I can see that PT takes an immediate dislike towards any volume function that is attached to the water.

I have managed to reduce the amount of noise when PT-rendering without the volume function. But the shadow as such is still a flat, disappointing blob with no transmission whatsoever! I don't know - maybe PT does not even require volume function?

1.jpg 2.jpg

One of the images has a higher decay distance, but that doesn't seem to change much. Contrary to the standard renderer which instantly reacts to the DD changes.

I'm a bit disappointed that Matt is not participating in this discussion, because sub-surface scattering and volumetric medium (apart from clouds) are seriously buggy in the path-tracing and require insane settings to get them right. Remember the point in time when we struggled to render a noise-free volumetric candle? There you go...

This needs to happen to the PTed water:

3.jpg
"This year - a factory of semiconductors. Next year - a factory of whole conductors!"

Tangled-Universe

Standard vs PT rendering of water can be considered like Cloud v2 vs Cloud v3.
The way cloud shader v3 "interprets" the fractal input is different than v2.
It's literally how density and scattering differ between the 2 shaders and how this results in a different look in your render.
To me there's a similar analogy between rendering water in standard and PT.
Without volumetrics reflections are already different, because the path tracer calculates these more precisely with (often) a softer result.
With volumetrics in your water the standard renderer does fake volumetric (plug in a PF and see for yourself how flat it looks) and PT does much more accurate (scatter towards normal) to even very realistic (scatter in all directions).

With that kind of explained, why you can not expect the same thing between the two...how to render...

In my experience rendering water with the PT is best with more paths per sample.
For instance, compare your results with AA4 and 25 max paths per sample and compare that to AA4 and 64 max paths per sample.
If you render with AA8 you already take up to 4x more primary samples compared to AA4, so you may need less max paths per sample than with what looked good/decent with AA4.
For water it's best (as in performance) to keep the volumetrics being calculated as "scatter towards normal".

N-drju

Thanks for the tips Martin but I found a nice workaround. 8) I'll definitely keep these suggestions in mind for future renders though.
"This year - a factory of semiconductors. Next year - a factory of whole conductors!"

WAS

I agree PT Water is probably more realistic, except the transmission. Shadows don't transmit through the density. This is something you can see in real life with muddy water and shadows. It's missing altogether in PT, from exmaples here and tests I've done. It exists in Standard. Something that needs work imo. It honestly looks as though the shadow is projected onto the SSS surface, and than just cuts through vertically.

Matt

Can you send me a TGD? The shadows will look different in PT (soft, and without the "rays"), but the hard edge isn't right.
Just because milk is white doesn't mean that clouds are made of milk.

N-drju

Quote from: Matt on June 16, 2020, 03:42:55 PMCan you send me a TGD? The shadows will look different in PT (soft, and without the "rays"), but the hard edge isn't right.

I can't. This is my contest entry (no offence).

How about a watered-down (no pun intended) version of it, after the contest is over?
"This year - a factory of semiconductors. Next year - a factory of whole conductors!"

WAS

I've been experimenting with this @Matt and I think it may be the AO/GISD like effect in dark shadows. Take this example. The actual shadows are fine, but the hard shadows at the banks are, well hard. Is this the terrain reflecting and just being really dark?

Matt

Quote from: N-drju on June 17, 2020, 02:15:07 AMI can't. This is my contest entry (no offence).

How about a watered-down (no pun intended) version of it, after the contest is over?


It's up to you, of course. No-one at Planetside is entering the contest or judging it, and any files we receive are kept confidential within the Planetside team.
Just because milk is white doesn't mean that clouds are made of milk.

Matt

Quote from: WAS on June 17, 2020, 04:56:39 PMI've been experimenting with this @Matt and I think it may be the AO/GISD like effect in dark shadows. Take this example. The actual shadows are fine, but the hard shadows at the banks are, well hard. Is this the terrain reflecting and just being really dark?

That's a reflection of the terrain. I think N-drju was talking about the shadow of the sun.
Just because milk is white doesn't mean that clouds are made of milk.