Welcome to Fractal Forums

Hi Jesse,

first of all a big THANK YOU for implementing the MC renderer. I don't want to moan about the well known tuning- and calculation-expenditure that comes with this method but ...

I really miss a percentage slider for the influence of secondary resp. indirect lighting. Gamma and exposure is not the same because it affects the whole image including ambient and direct lighting.

In this example i tried to illuminate the scene by secondary light coming from the floor. But it doesn't work the way i imagined that. Is there a way to regulate the strength of GI?

(http://www.feelgreat.at/lxh/temp/mc_test.jpg)

Hi lxh,

it is possible to enhance the ambient occlusion by the settings of:
'Ambient rays depth', that means how many diffuse reflections in a row (meaning a random light reflection is reflected again and again... ) are calculated

and by the length of those rays that are calculated at maximum, that is given by 'MaxL' in the AO postprocessing tab for the DEAO function!
You can send the parameters from MC to main, then change the value there, and import back.

Of course render time depends strongly on these settings, check if you can increase the raystep multiplier.

Also the intern gamma setting (I2 on the lighting tab) and the diffuse colors itself got an influence of the ambient strength.
A lighter diffuse color reflects more light...

good luck

ps, does not look that bad the way it is

Ah, yeah, got it. Wow ... :D
I never thought of 'damped' rays by DEAO. Man, i think no one here did.
Thanks a lot, Jesse!

Hmm, spending houndreds of hours for all the functualities and saving 1 minute for not mentioning this in the readme,
seems a good compromise for me  :D

Hehe, this is the keyword. Saving time and compromise:

A higher value of MaxL brings up longer rays and as consequence more secondary light but it seems that it kills the ambient light proportionally.

As you can see here (MaxL 10 with dramatically increased calculation time), the foreground has a nicer diffuse 'ground light' but the background lost its ambiance.

(http://www.feelgreat.at/lxh/temp/mc_test2.jpg)

Well, in this case i've made a lazy compromise and layered these two images plain but ...

(http://www.feelgreat.at/lxh/temp/mc_test3.jpg)

Is there a way to get longer rays AND ambient light in one single rendering?
Or .. an optional ZBuffer output from the MC renderer could help to mix it the _right_ way.

Anyway ... please forget my question, Jesse.
MaxL 10 is a little bit too high an not really necessary. Beside that i can screw up the ambient light.
But tuning MC is killer. And i'd better never seen Octane renderer's workflow and performance. :-\

I did not see, but I saw the specifications. Looking good at a surprisingly moderate price...
But to compare this with m3d is not really fair :dink:

Yeah, you are right Torsten. And i am sorry for that. But i just saw ...

http://www.youtube.com/watch?v=CZ1IRQTqMMY (http://www.youtube.com/watch?v=CZ1IRQTqMMY)

... and other impressive demonstrations and i seriously consider to invest in some hard- and software. To bad that there is no way around voxelstack at the moment, but this is where the journey goes.

You've always got the option of post-processing your renders in a tonemapping application such as Picturenaut or Luminance HDR (both free).

Good luck! Only the rendering Machine used in this demo has the exchange value of a compact car. Not to mention the energy costs...

But seriously, the most impressive thing with those renderers is the interactivity. I never cared about final rendering time of some hours, at least for still-frames. But a realtime preview of that quality really kicks!


I'm not up to date. With 48bit pngs this would really make sense. Does m3d produce pngs of that depth?

Thanks for the input blob. HDR might be an option in this case and i'll give it a try. But i will prevent multiple mc renderings from now on. :)

No no no, hehe. I won't sell my car to buy a computer. ;D
But one good single gpu like nvidia titan in my comp should do a good job.
And yes! It's all about tuning. Tuning lights and materials in nearly realtime. Not the final rendertime is the prob ...

Beware that this demo was probably rigged in some way. Nothing major, no fakes and such, but I strongly suspect we are being shown a best case. This is advertisement after all.

For example, why did they turn motion blur off? True, it wasn't really needed for such a static scene. But don't you think they would have bragged about being able to do motion blur as well, if only they actually could do that with perfect performance scaling, too?

There is probably a bottleneck updating the scene for different points in time during rendering. So before you spend a mountain of money, try to get a hands-on demo specifically for you, for your scenes and your workflow.

Thanks a lot for your warning, hobold. I know what you mean and i will not run out right now spending a mountain of money. ;)
Sooner or later every budget gpu can do this. The challenge is to find a good and economical method to let fractal images benefit of this technology.

Important point. a radeon 7870 or a GTX660 has about 70% of the titan perfomance but costs 1/5. When you're using it to earn money and every minute costs, take the big one. In every other case, i'd try to find a suitable compromise.

I don't think M3D outputs 48bit pngs but I might be wrong and failed to see the option. Still worth a shot with LDR pngs IMO even though it's not optimal you still can get interesting/decent/suitable results.

Well, that's a point (and totally off topic now ;D). But radeon gpu's don't have CUDA and a GTX660 doesn't have 6GB ram. Maybe i can get a 660 with exchange option.

Yes. In this case it's not about outrageous dynamic ranges but a simple 'combining' problem.  If i just want to mix the light parts of the second with the darker parts of the first, hdr is a good idea and has no need for higher bit resolutions. Thanks again.

Hmm, the lost in ambient can't be due to the ambient raylength itself, but maybe if you enabled dynamic fog and light is subtracted!
Check the 'dont subtract light' in the dFog options if you got some amount of fog.



You can send the parameters to main, render a fast image and save the Zbuffer then.
For antialiasing you might increase the imagesize by 2 before.

Thanks for your answers, Jesse. I've tested that. With or without dynamic fog or any combination of its options, it has no effect to this phenomenon.
It seems the correlation of light-ray length and shadow-ray length does a good job for the ambient occlusion but work against ambient lights which i screwed up to the maximum while lowering the ambient shadow to the bottom end. I've also tested it without L2 Gamma. It stays the same. The DEAO shadow at higher MaxL values is simply too massive so that it destroys any ambiance inside this scene while increasing the ray length for the ambient occlusion. Well, as i understand it now, i can name it a ratio-regulation between fake- and real ambient light and would - if i could - separate the light-ray length from the shadow-ray length.

But another thought ... is it just because i use v1.88 and should i get the latest version in this context?


That was my idea too but i thought the anti aliasing wouldn't be exactly the same and would lead to some edges. But i will test it ...
Thanks again, Jesse!

Only a few settings does have at all an effect on the mc renderer, there is no ambient shadow that is calculated.. only light will be added from all diffuse rays send out.  So also the ambient shadow settings in the lighting tab have no effect.
The only ways of subtracting light on the way is the dynamic and the depth fog, maybe you got dark depth colors and the depth slider is not at zero position?



A higher MaxL value can indeed lead to darker ambience when walls are hidden that are dark like in this case.
If no object was found until MaxL the background light is added, so this might be the case why it will be darker, but this would be the expected case.
You would have to increase the ambient rays depth count and lighten the diffuse colors maybe.


Should be no difference for this case.

[/quote]
That was my idea too but i thought the anti aliasing wouldn't be exactly the same and would lead to some edges. But i will test it ...
Thanks again, Jesse!
[/quote]

Hmm, i have not tested it yet, but with some luck it should fit well.

A really fractal cathedral  :wow:, it seem a reproduced of the palace "Worldstone Keep" from game DIABLO 2 whick have same shape.

Yes, i see. I checked it right away my posting. Sent it back and realized the reseted ambient shadow. But the normal fog is totally turned off and black.


Hmmm, i see the functionality and it seems logical to me but it doesn't change the fact that i still miss the option to mix longer rays with faked ambient lights. Well, this case might be an exception but i recognized the same tendency in other - more opened - scenes. Normally i wouldn't say a word because of the option to do multi renderings and layer them. But if you wait .. say 40 hours for a real good high resolution image, this circumstance becomes a theme.

Thanks for your support, Jesse!


Thanks a lot hgjf2! :happy:

PS:

I might be up to a trick that could help. In this example we have a pretty similar light situation where i tried to increase the mc intensity by adding diffuse reflections. First test with an amount of 0.5 and mc reflection diffusivity of 1.0, depths 1. It seems that diffuse reflections reach out much further so there's no need to increase the MaxL value and loose the backgrounds ambient light. Well, it gives a little different look .. actually it increases details like a halved DEStop value. Maybe a higher diffuse value can make it smoother ... i will test it soon.

So if there's no other need for reflections, it could be an option to misuse it by setting high diffuse values.

(http://www.feelgreat.at/lxh/temp/mc_test5.jpg)

Hi Jesse,

I know what the problem was. I had a simply too dark texture mapped by color wrap function. It damped the secondary light.
So i just made this texture brighter with less contrast. Here's the result ...

(http://www.feelgreat.at/lxh/temp/mc_test6.jpg)

But this was too much for my taste and i mixed these images over 2xAA zBuffer from the normal renderer which seems to work properly with MC images.

(http://www.feelgreat.at/lxh/temp/mc_test7.jpg)

Thank you for your efforts.

Yes, that looks very nice  :)

That was what i meant with a lighter diffuse color, though my explanations were not that clear i fear.
Was not that sure by myself...
so we need no fake ambient shadows and i don't have to say that i never do fakes into this renderer, but try to get rid of the compromises it still has  :dink:

Thank you Jesse. :)

The definition of fake? I think every intervention to the original render output like photoshopping a zBuffer map or tuning colors and contrasts is the same 'fake' like changing the ambient light color or ray length in the pre-render tuning procedure. So i suppose rendering itself is the actual fake. ;)

But the real problem is tuning the mc renderer. Do you think it's possible to drill up the mc function and do some internal zBuffer and normalmap masking and layering like you did in the standard renderer? Internal multipass might be the right therm. Well, i know it isn't possible to change (for example) the highlight specular value on the fly because it has to be calculated slowly within the whole mc render process but it could be possible to layer it internally and change its opacity or do even a gamma correction to fake its specularity afterwards. The same with reflections for example. OK, it could look more and more like a fake if then someone tunes such layers in extreme directions where it stands out that the whole lighting impression can't be true. But tuning mc has more of a subtle affair where small changes don't look faked but make the big difference.