How to go about the new SSS option?

[N-drju]

I'm surprised it's still so noisy. Can we test your TGD and object to understand why this is happening?

Of course you can! But there is not much of a .tgd in this case - just a default scene with sun lowered under the horizon.

I attach a different object which is sculpted into something that is closer to what the final candle will look like. Let's go from here, okay?

I am also confused by the fact that the light in the final picture is apparently handled in a an absurd way. Just take a look at the picture - while the waxy surface is being lit up much better than the few last times, the "burnout rim" of the candle stays completely black... And this is all just a single, uniform object...

No message is associated with this attachment.

[/attach]montownia (gathered).tgd<br><br>swiczka.tgo

[WAS]

AA 10, Max Paths....

Edit: I emailed you the scene Matt. I couldn't find the candle again on Sketchfab to credit the maker so couldn't upload here.

[N-drju]

Still grainy, but rest assured, your effect is great!

[WAS]

Still grainy, but rest assured, your effect is great!

If you PM me your email I'll let you have the scene too so you can inspect the SSS settings and lighting.

[N-drju]

Here is one more render with a "fake" sub surface scattering, Hannes - style (an object within an object). Fast, simple, effect clearly visible.

Dirty as clean. But the more reason to ask why do we need an imperfect glass shader to achieve the same? Please note that this render was path traced with AA=5 and is still less grainy than glass shader at AA=10!

[Matt]

Thanks for the downloads, N-drju. I've been doing some tests with your files and discovered some things which might help.

To reduce noise, reset transparency to 1 and reduce the lightsource intensity. Transparency controls the surface transparency, not the transparency of the whole object. Before the subsurface scattering can do anything, light has to pass through the surface, so 1 is a good value for this. Increasing transparency makes the object much brighter so now you need to reduce your light intensity. Your scene had light source intensity 6 and I reduced it to 1. Now it renders with less noise.

I am also confused by the fact that the light in the final picture is apparently handled in a an absurd way. Just take a look at the picture - while the waxy surface is being lit up much better than the few last times, the "burnout rim" of the candle stays completely black... And this is all just a single, uniform object...

This can be improved by changing Lighting Method in PT to "Subsurface scatter in all directions (BETA)". There will still be a sharp line where the candle transitions from the thin rim to the flat section in the middle, but it can be improved by making some more changes.

Increase the specular roughness to 0.8. This scatters the rays entering the wax. It also affects the specular reflections though, so if you need additional sheen on the outer surface you can follow this shader with another reflective shader.

To soften this edge even more, I think you'll need to create wider rim and perhaps more rounded at the bottom. I would also try to round off the edge of the rim because the razor thinness causes the edge to be too transparent. Imagine it is smokey glass that comes to a razor thin edge - this will not scatter any light (EDIT: actually this isn't so bad after making other changes to the shader).

[Matt]

Increasing decay distance to 3 and volume density to 10 helps to brighten the material and I think this works better.

[WAS]

Uhmm, what about the biggest concern evn if he (or I) got good settings, which is quality? AA10 is absurd, for a noisy render that looks that bad.

Not to be a jerk, but I also find it convenient you ignore my input (and file submitted), as well as the obvious quality concern in my scene with working SSS. This seems to be the unavoidable part, and isn't even addressed in your replies.

I even have an entire topic regarding it in support with no reply.

One thing I found helps is lowering the pixel noise threshold and sample size, but this hits render time like a ton of bricks. I really feel more paths would help this considering more lit areas being handled well, even under lower AA>

[N-drju]

Hey Matt. I'm glad you researched my file, it's very kind of you.

I will sure take these tips into consideration. I must admit that transparency measure and decay is somewhat confusing. Especially when someone, like me, does not usually use such shaders extensively!

There will still be a sharp line where the candle transitions from the thin rim to the flat section in the middle, but it can be improved by making some more changes.

To be frank, this is something that I have noticed not long after I uploaded the files... The way an object is modeled, is very important indeed so one needs to be careful when moddeling... Placeholder objects will render poorly.

Despite all you have said, I still believe that there is some significant, not entirely clear "magic" going on if Bastien is able to render some "wowsy" renders at apparently no cost. While all I get is a noisy render. In fact, so noisy that the final effect is not even worth trying...

I'd like you to consider the renders I have made with the fake SSS, mentioned before. I can make such renders at a fraction of the time and AA value that I would otherwise have to deal with when rendering a glass-based SSS... And the effect is quite pleasant, don't you think? These were PTed at 0.5 MPD, 6AA, 81 P/sample:

[Matt]

Not to be a jerk, but I also find it convenient you ignore my input (and file submitted), as well as the obvious quality concern in my scene with working SSS. This seems to be the unavoidable part, and isn't even addressed in your replies.

N-drju's scene appeared to have more problems than yours, so I thought I could do more good with his scene and I focused on that first. It took a considerable chunk of my day to work on this, and it's not the only part of my job. Terragen doesn't grow on trees. It's heartwarming to know that my efforts are appreciated.

Uhmm, what about the biggest concern evn if he (or I) got good settings, which is quality? AA10 is absurd, for a noisy render that looks that bad.

I even have an entire topic regarding it in support with no reply.

I found that with N-drju's scene, after making the suggested changes the noise was quite acceptable in an AA7 render. I hoped they might translate over to your scene as well, but I have not had time to try it yet.

One thing I found helps is lowering the pixel noise threshold and sample size, but this hits render time like a ton of bricks. I really feel more paths would help this considering more lit areas being handled well, even under lower AA>

Yes, I think more paths will help. I am considering ways to add this to the GUI in future versions.

[Matt]

Despite all you have said, I still believe that there is some significant, not entirely clear "magic" going on if Bastien is able to render some "wowsy" renders at apparently no cost. While all I get is a noisy render. In fact, so noisy that the final effect is not even worth trying...

Resetting transparency to 1 should make a big difference to the noise. Please give it a try. But yeah, some scenes are still going to be noisy and difficult to render. I don't know for sure, but Bastien's renders probably had AA much higher than 10. I may be remembering wrong, but I think some of his renders were AA 16.

We need to write some guides to SSS, I appreciate you bringing this up.

I'd like you to consider the renders I have made with the fake SSS, mentioned before. I can make such renders at a fraction of the time and AA value that I would otherwise have to deal with when rendering a glass-based SSS... And the effect is quite pleasant, don't you think? These were PTed at 0.5 MPD, 6AA, 81 P/sample:[/font][/size]

Sure, it is simpler to compute, and if you're happy with these results then that's great. There are diminishing returns with some more computationally intensitve algorithms. Sometimes the results are worth the cost, sometimes they are not. In many cases there are multiple ways to achieve the same effects, and if you've got a faster way, use it.

The path tracer and SSS in Terragen are not "done", they are going to see improvements over time and hopefully we'll get much faster and cleaner renders in future. The renders you're doing are showing areas where I should focus my efforts, they've definitely got my attention and I'll do what I can to make the renderer better.

[Matt]

Whenever you're testing scenes where the background doesn't matter much, disabling the planet and background sphere will speed up your PT renders quite a lot. Even if you only disable them for test renders, it's worth doing it to speed up experiments.

[WAS]

SSS guides, especially focusing on scale (so that becomes habit to the user) would be great. I think a lot of confusion comes from scale of objects. I notice the decay distance could maybe use some refining. WHen I have an object 2m, and have a decay of 1m, things seem to act as if it was a much lower decay distance. Least to my eyes.

I think I've noticed this with just the water shader in the past. Having lets say a 5m depth pond, and a 5m decay distance doesn't seem right and you often gotta double that up.

[Matt]

Yes, scale is important. It's more difficult to translate materials from one object to another if the objects scales are different, but luckily there is only one parameter that you need to change to accommodate incorrect object scale, and that's the "decay distance".

[WAS]

Yeah, but take a simple sphere for example.

A 1m sphere, with a 1m decay distance, in relation to other engines, and just natural thought on the process, this sphere should look like a creamy piece of marble or something. But it's very dark, and doesn't let much light through at 0.25m, let alone 1m.


This may be leading people to setup up their SSS incorrectly based on their judgement from the named settings.