Low lit room issues

[Maximus]

I'm not sure what you trying to do but if i got it right you want to mix two kind of dirts in the wall using a black and white image as the mixer driver.
In this case you should plug the b/w image into the weight port of the mix2color node or any general mixing nodes like here:



Default black is base color while white is color 1, of course you can invert the B/w image to swap those inputs.

[steven07]

Yes! Exactly! Thanks

[Falam]

Those are FG splotches, you get rid of them by increasing the number of rays, point interpolation or filter size.

FG in an interior low lit or indirect lighting scene is really hard to tweak, but you have a fair amount of light in you scene so it should do it.

-Decrease the presample density to a minimum level, and concentrate on rays and point interpolation, something like 0.25 to 0.4 for presample density (depends on the complexity of your geometry in the scene)

-250 to 450 FG Rays, sometime it can go in the thousands. Point interpolation is a areal player, it can get rid of splotches real quick at the expense of realism, start from 15 to 50 or more. interpolation work in conjunction with FG rays, more FG rays less point interpolation need it, more realistic result, yes you get it. The real spoiled Princess is FG rays.

-If you are looking for good indirect lighting with FG forget filter it discard fg rays that can cause splotches, unfortunately it eat indirect illumination a lot. so put it at Zero if you want, and use FG rays and point interpolation.

-Ambient occlusion is important if you want to bring back the detail that you scarified in lowering the presample density, activate it in every materials you see that need it. increase the shadow color the get softer look.

-Unified Sampling (US) it's a long story and I'm still at half the book, I'll give you the quick resume: US is clearly superior to Adaptive subdivision (AS) in mental ray. If you had sampling problem with glossy reflection, direct specular reflection in the past with AS, US get rid of it beautifully, it just cut trough everything, and sometimes it faster than AS, as a say "sometimes" because AS is still faster in some scenario like your scene with simple diffuse lighting low glossy reflection, and simple geometry, but when you start lowering your glossy reflection, refraction, adding fine detail geometry and high direct and direct light specular, DOF. just switch to Unified is just easy.

You start at min of 1 and max to big numbers like hundreds of samples, depends on your scene, and you control your result with quality slider, when it stop having effect and you still not happy with the sampling, you know you should increase the max sample.

-Some sampling numbers for your materials and lights if you are using US.

Glossy refection: 1 to 12, why ? Sometimes you get rid of noise with just 1 sample, but Unified is still not smart enough to know where to concentrate it sampling and energy. 1 sample will get rid of your noise sometimes, but also of your light energy so you need to help the sampling a bit by increasing the glossy reflection

Area Lights: 1 to 8, same music, like glossy reflection.

AO: 1 to 6 or 12, this one I'm not clearly sure, I just put it to 4 a leave it be.

Well that's it I think. Good luck. Don't hesitate to ask more questions.

Do you find these settings will speed up a render ? I have all sampling for the Arch shader at zero, and I'm letting US take care of it, so far I can't say that I've noticed that much of a difference compared to AS.

[Kzin]

for glossis, use minimum of 2 samples. why? because of the bias the mia mat generates with low samples. the more samples, the more detail has the reflection, or lets say the more correct mia samples the surrounding objects. 0 samples for glossis will end up with no glossis at all. for ultra wide glossis you could easily end up with 32 samples for smooth results, means values like 0.25 or lower.

a good alternative in combination with ibl light is to use highlight mode only in mia mat. because the highlight of the ibl is like a reflection of the environment you get glossis of it without additionel rendertime.

[Falam]

for glossis, use minimum of 2 samples. why? because of the bias the mia mat generates with low samples. the more samples, the more detail has the reflection, or lets say the more correct mia samples the surrounding objects. 0 samples for glossis will end up with no glossis at all. for ultra wide glossis you could easily end up with 32 samples for smooth results, means values like 0.25 or lower.

a good alternative in combination with ibl light is to use highlight mode only in mia mat. because the highlight of the ibl is like a reflection of the environment you get glossis of it without additionel rendertime.

That is strictly AS settings within the shader. What about US ? If I set all my gloss samling to 2, and I have my US sampling at 1/100 and the cutoff at .003 which I read is reasonable. There is another setting in between, I forgot I'm rendering the image now so I can't access

[Kzin]

That is strictly AS settings within the shader. What about US ? If I set all my gloss samling to 2, and I have my US sampling at 1/100 and the cutoff at .003 which I read is reasonable. There is another setting in between, I forgot I'm rendering the image now so I can't access

my comments are always for unified because i dont use adaptive anymore.
the 2 sample minimum for glossi rays comes from my own observation and alot of tests i run. you can use the 1 sample method, but because of the bias thing you should use higher reflection intensitys.
and dont take every setting as nailed in stone.
min1/max100, q of 3 is a good STARTING POINT with alot of arealights and glossis, nothing more. its possible that you need 300 for max or sometimes min1/max64/q4 can be enough. it depends on the scene, geometry and shaders.

[Falam]

That is strictly AS settings within the shader. What about US ? If I set all my gloss samling to 2, and I have my US sampling at 1/100 and the cutoff at .003 which I read is reasonable. There is another setting in between, I forgot I'm rendering the image now so I can't access

my comments are always for unified because i dont use adaptive anymore.
the 2 sample minimum for glossi rays comes from my own observation and alot of tests i run. you can use the 1 sample method, but .
and dont take every setting as nailed in stone.
min1/max100, q of 3 is a good STARTING POINT with alot of arealights and glossis, nothing more. its possible that you need 300 for max or sometimes min1/max64/q4 can be enough. it depends on the scene, geometry and shaders.

The 2 sample, you are referring to AS within the shader, correct ? What do you mean "because of the bias thing you should use higher reflection intensitys" ?

[Kzin]

The 2 sample, you are referring to AS within the shader, correct ? What do you mean "because of the bias thing you should use higher reflection intensitys" ?

i dont mean as, i mean unified. try to render a glossi surface with adaptive and 2 local glossi samples, it will not work. and i wrote that i always use unified. ;)
mia does not sample all details with low local glossi samples when using glossi reflections. the more local rays you use in the mia mat, the more correct will be the reflection. for example, if you have a bright object in your scene and wants to get all reflectiondetails from it on your detailed model, you need much more local glossi rays then 2. its just a accuracy problem in the current mia mat. it cant solve all the details without more rays. the bsdf shaders dont have these problems for example. i think because they use importance to sample the scene they are more correct by default. the problem itself is only one when you are sensitive to tiny, little details which no one will see in an animation. ;)

[Falam]

The 2 sample, you are referring to AS within the shader, correct ? What do you mean "because of the bias thing you should use higher reflection intensitys" ?

i dont mean as, i mean unified. try to render a glossi surface with adaptive and 2 local glossi samples, it will not work. and i wrote that i always use unified. ;)
mia does not sample all details with low local glossi samples when using glossi reflections. the more local rays you use in the mia mat, the more correct will be the reflection. for example, if you have a bright object in your scene and wants to get all reflectiondetails from it on your detailed model, you need much more local glossi rays then 2. its just a accuracy problem in the current mia mat. it cant solve all the details without more rays. the bsdf shaders dont have these problems for example. i think because they use importance to sample the scene they are more correct by default. the problem itself is only one when you are sensitive to tiny, little details which no one will see in an animation. ;)

Within the Arch/Mia shader set the glossy to "2" and the rest Unified ? That is the message I'm getting if you want quite accurate reflections. I have a very highly gloss reflective interior shot, so far after :40 minutes at roughly, just 640 resolution the render region is still not done, at 720 or 1080p I could be looking at 10 hrs, and this scene will be rendered on a fast PC, is there any suggestions, for not this this project but any in the future when using the Arch shader ? I am using FG and GI in the scene, but I have adjusted all the settings including number of rays etc to eliminate splotches etc. Number of rays in FG was at 22, now in this test to final render I crank it up to roughly 450.

[Kzin]

The 2 sample, you are referring to AS within the shader, correct ? What do you mean "because of the bias thing you should use higher reflection intensitys" ?

i dont mean as, i mean unified. try to render a glossi surface with adaptive and 2 local glossi samples, it will not work. and i wrote that i always use unified. ;)
mia does not sample all details with low local glossi samples when using glossi reflections. the more local rays you use in the mia mat, the more correct will be the reflection. for example, if you have a bright object in your scene and wants to get all reflectiondetails from it on your detailed model, you need much more local glossi rays then 2. its just a accuracy problem in the current mia mat. it cant solve all the details without more rays. the bsdf shaders dont have these problems for example. i think because they use importance to sample the scene they are more correct by default. the problem itself is only one when you are sensitive to tiny, little details which no one will see in an animation. ;)

Within the Arch/Mia shader set the glossy to "2" and the rest Unified ? That is the message I'm getting if you want quite accurate reflections. I have a very highly gloss reflective interior shot, so far after :40 minutes at roughly, just 640 resolution the render region is still not done, at 720 or 1080p I could be looking at 10 hrs, and this scene will be rendered on a fast PC, is there any suggestions, for not this this project but any in the future when using the Arch shader ? I am using FG and GI in the scene, but I have adjusted all the settings including number of rays etc to eliminate splotches etc. Number of rays in FG was at 22, now in this test to final render I crank it up to roughly 450.

a fully glossi room?
can you post your ray stats? it stand at the end of your verbose messages.
my first advice would be to use only the raydepth that is visually most important. another would be a higher ray treshold in the mia mat to exclude details which no one will see.
have you tryed the bsdf shader for reflections? they are twice as fast compared to mia in most scenes minimum.

[Falam]

a fully glossi room?
can you post your ray stats? it stand at the end of your verbose messages.
my first advice would be to use only the raydepth that is visually most important. another would be a higher ray treshold in the mia mat to exclude details which no one will see.
have you tryed the bsdf shader for reflections? they are twice as fast compared to mia in most scenes minimum.

The room is not fully glossy, it's glossy with alot of reflections, like a metallic room.

I'll post my ray stats after this render region is finished, still not complete Do you have any suggestions to speed it up as the scene currently is ? This is going on 2.5 hrs

[steven07]

Question: should everything I use be the mia_material_x node? or just reflective / refractive objects?

[Kzin]

Question: should everything I use be the mia_material_x node? or just reflective / refractive objects?

also as soon as you use indirect light, mia is best here.

[steven07]

Well let me ask about the other extreme then. Say I have a studio lighting animation where everything is just infinite white and I have objects moving around and deforming. Would I get flicker with final gather? I know LightWave's radiosity get's a lot of flicker without cache.. and it doesn't even work with moving objects..
Is MR similar in that sense? Or is that something that a renderer like VRay would be better at?

[Kzin]

Well let me ask about the other extreme then. Say I have a studio lighting animation where everything is just infinite white and I have objects moving around and deforming. Would I get flicker with final gather? I know LightWave's radiosity get's a lot of flicker without cache.. and it doesn't even work with moving objects..
Is MR similar in that sense? Or is that something that a renderer like VRay would be better at?

you can get flickerfree gi in mr. from my experience mr is better then vray in flickerfree animations for interpolated gi with more indirect light.
but you have to take care about for some thing. you need a good amount of direct light as a good base for fg. dont try to solve lighting details with fg. did alot of tests and its better to use direct light together with low fg setting and, if you need it for detail light details, the ao color bleed feature with short distance for a stable solution. this will also work for character animation.

i dont know why everyone thinks vray is better here. it could be better in shots with alot of directlights, yep. its better if you rely on brute force, but only if you dont have alot of indirect light. in this case, its as slow as mr's path trace feature. alot of indirect light cant be solved with ir and lc without extremely high rendertimes. they are that high that is does not make any sense to rely on interpolation, brute force is not slower. and alot of companys are baking the gi in vray, the old background/forground character trick.

[RichardCulver]

Well let me ask about the other extreme then. Say I have a studio lighting animation where everything is just infinite white and I have objects moving around and deforming. Would I get flicker with final gather? I know LightWave's radiosity get's a lot of flicker without cache.. and it doesn't even work with moving objects..
Is MR similar in that sense? Or is that something that a renderer like VRay would be better at?

Vray, LightWave, Arnold (As it is in Messiah Studio) and Mental Ray all have a "similar" approach to rendering GI. 3Delight is another animal but it has also similar problems.

The basic problem to put it very simply across all of these render solutions, is a matter of having enough rays to gather enough information so that there is little or no change between frames and to capture detail.

The basic concept of interpolation:

http://en.wikipedia.org/wiki/Interpolation
Newcomer (<20 posts) alert: please use the URL tags - HB

In the mathematical field of numerical analysis, interpolation is a method of constructing new data points within the range of a discrete set of known data points.

In engineering and science, one often has a number of data points, obtained by sampling or experimentation, which represent the values of a function for a limited number of values of the independent variable. It is often required to interpolate (i.e. estimate) the value of that function for an intermediate value of the independent variable. This may be achieved by curve fitting or regression analysis.

A different problem which is closely related to interpolation is the approximation of a complicated function by a simple function. Suppose the formula for some given function is known, but too complex to evaluate efficiently. A few known data points from the original function can be used to create an interpolation based on a simpler function. Of course, when a simple function is used to estimate data points from the original, interpolation errors are usually present; however, depending on the problem domain and the interpolation method used, the gain in simplicity may be of greater value than the resultant loss in accuracy.

So what is that saying basically? It means that in order to get a faster render (in this application) you sacrifice accuracy. This is noticeable in a moving shot where the camera moves, even more noticeable when you move an object and further exasperated when you move meshes on the point level, or have detailed rendering of hair or particles and bumpmaps/displacement maps with a lot of detail.

This is because you are only taking a fewer set of sample points and estimating what is in between these points, rather than taking direct samples of the points in between. This is over simplified. But it is the basic problem. And all of the methods - though they vary in approach and math and so on - are basically doing one of 2 things. They are either sampling everything or taking lesser samples and interpolating. And of course and any combination of the two.

The solutions to the problems of taking fewer samples are in caching and point clouds and so on. These solutions have limitations. As you are aware of in LightWave. Or to tweak the interpolation settings and brighten up the lights and so on. To me these things make faster renders, but take the hit on the finer artistic elements I am looking for in a lit shot.

My take on it personally, is not to spend countless hours tweaking lights, making sure everything is bright enough not to flicker. And tweaking all of the various interpolation settings. (Usually toward sampling more and more points anyway which takes longer) But rather, to use non-interpolated methods. This is killer on interiors but much much faster on outdoor shots.

With non-interpolated methods basically refereed to as "brute force" (available in all of the software mentioned above) you have much more freedom with your lighting schemes for interiors when you are going to be animating - especially with characters. And you can bring out the finer details because things are not being glossed over and interpolated. To me it is like night and day. No pun intended.

That is what several months of testing, training and using all of the above mentioned software taught me. That is, for my artistic sensibilities. Many of the other "solutions" I see come at the sacrifice of the look I am going for.

And I do know that is not for everyone. (Or every situation for that matter practically speaking) But just thought I'd share this insight and you can make your your own choices based on your needs.

The other option is to just become damn good at lighting and fake it. But, the problem here is it will be very difficult to really do good GI fakes that hold up to the real thing.

In the end there is no absolute solution. They all come at a sacrifice. It is just a matter of what are you willing to sacrifice? And for each artist, that will be something different.

That's my take.