Mental Ray Doubts

[Undseth]

If you are using maya with mentalray 3.4, I think you should try increading the max final gathering radius, even if this isnt the way you used to do it with earlier versions of mentalray.

[Droolz]

I found MR frustrating, my alternative was to look at different renderers...

http://www.cgtalk.com/showthread.php?t=255010

[gga]

1. Jitter in animation if its normal maya shaders or mental ray shaders,
settings are very high at production Quality, but i still get jitter at a final gather of 5000 even some cases 15000


If you are getting jitter with final gather at 5000, you are doing something wrong, like your radii are incorrect for your scene. In maya 6.5, you can also start using much lower values for the same quality as before (usually not having to go beyond 1000).


2. the geometries are pretty heavy and i use raytrace shadows which means high render times like 5-8 minutes a frame.


5 minutes a frame is usually considered low render times. In general, for tv size renders, up to 15 minutes is considered acceptable.


4. layer rendering is another option we tried but its too time consuming since the files take 2 minutes to open each time and are hi poly meshes for episodical work.


You are only opening a single file once and then spitting out all layers, right? If not, you do not understand what layer rendering is.
You mention episodical work. If you are dealing with a 30 min. tv series, truth is that you should be looking at keeping your render frames at 3 mins or less.
In that case, any sort of global illumination of any renderer will simply be too slow for you. You should stick to scanline rendering mainly and use raytracing also sporadically.
In a renderer like mental ray, you should also try to build your geometry with nurbs or meshes with subd approximation, and let the renderer handle the tesselation at render time (ie. not use things like polysmooth and similar, which will slow both the rendering and the opening of the scene file).

[JasonA]

wow, I must be doing something really wrong then, nbecasue I have rendertimes that go waaaaay beyond yours... :banghead: So I have to ask, but how many poly's are in your scene??

I'm working on doing FG rendering (with mr 3.3) of my mclaren model, and my render times are like 2 hours a frame.. my scene density is about 800k poly's, I'm using Mitchell filter 0/2, and IBL/FG. I'm only shooting 200FG rays. I'm using the standard 10%/1% for radii right now and I have pre-calucated (not rebuilding) my FG map. I'm using FG falloff about half the length of my car. My IBL light emission quality is 16/8. I thought I had my scene optimized as best I could, but my renders take foooreeeevvverrr... So can someone tell me what I'm missing??

btw I'm running on dual 2.66GHz Xeons with 2GB ram.

[Undseth]

Could it be due to you using lights from the IBL?;
"My IBL light emission quality is 16/8."

I have never used that thing for emitting light that way, isn't this creating lights which each cast shadows? Heh or am I wrong?

Seems like you can throw a lot of things (shadows, diffuse reflections etc) into the final gathering thingy, but it could take a loong time to do the calculations.

[BillSpradlin]

Using the "Emit Light" feature from the IBL node will murder your render times. It's basically creating area lights based off the texture map/texture that's plugged into it. If you have lights already in the scene there's really no need to use this at all.

[JasonA]

yes, I am using the ibl for lights as well as FG. I was doing this mainly because I got a much better final image. I have no other lights in the scene right now, but perhaps I need to review my setup. I mainly followed the 'mr masterclass' notes by paolo berto for my setup and dialing the scene in.

[BillSpradlin]

You are better off creating your own lights in the scene, area lights are extremely expensive to use.

[JasonA]

ah yeah even the masterclass notes mentioned this.. .. ok, I'll setup my own light sources and give it another whirl. Thanks for reminding me about this guys, I should've realized the Emit Lights was killing me.. :D

[danielkenobi]

Hi guys could you explain what is the IBL thing that you are talking about. My mental ray renders are time consuming and in some cases are very simple, right now I am rendering a scene wich is only one room with 8 spotlights using global ilumination and it takes about 10 minuts each render :sad: I am doing something wrong, also do u know if there is some kind of exposhure control so I can easily tune my final image?
Also does u have any ideas why the shadow of the ball loks like pixelated?
http://kenobi.pixeldeck.com/cgtalk/Varios/cuarto.jpg

[Undseth]

I can tell you about that shadow from the ball. You are probably using final gathering. And your final gathering (FG) min radius isn't small enough.

The IBL (Image Based Lighting), can be created from the render globals. It shows up in the viewport as a cage which you can rotate (according to P.Berto, you can only rotate it, other manipulations doesn't really matter if I remember right). You can give the cage a color, to use with FG, instead of using the environment color in the perspective camera, casting light when using FG. Or you can texture the IBL cage with a hdri image, casting light with FG this way. The IBL can also emit light, in other ways.

I have been wondering about the use of Global Illumination (GI) and FG, and I think I finally figured it out so to say. I think you get nice results if you GI radius is really small, and FG radius is large(r) (I tried x1 x10 x100 for GI radius, fg minradius and fgmaxradius). However I think one needs a LOT of photons this way. My machine can only cope with about 12 mill photons. Im no expert, and I really dont know for sure, but could it be that one really needs massive amounts of photons to really make a fantastic image. I never figured out the FG/GI thingy before, so it could be that I was just happy to get any decent result when I tried it out the last time.

[kapollo]

Hi there!

about fg min/max radii
as far as I know mental ray's unit mesurement is expressed in milimeters. But in maya one can operate on other mesurements (inch, meter, etc.)
Now does the mayatomr translator convert those maya's units to mr units ?
If not then the min/max radii should be first converted to mm.

As far as I remeber in the miiss_fast_something shaders one had to put inputs in mm.
But I wouldn't put my head on a bet that this is true.

@ danielkenobi:
I think You should play a little bit with constrast settings (if it wont help increase sampling a bit) because that shadow looks a little bit undersampled to me.
To tune the solution way the exposure would do you can play with photon intensity.

cheers kapollo

PS If I made an language mistake, sooooorrrry english isn't my native language.
Oh and have a good summer break everyone. bye!

[LehaS]

as far as I know mental ray's unit mesurement is expressed in milimeters

Well ...I know the same about centimetres...

Since the default unit for MAYA is centimeter i assume MR should use them

[Karlfucious]

Personally I dont even use final gather or gi anymore unless its really needed. I have had much better luck using the dirt map shader instead. Not only does it seem to be a lot faster than final gather it is a lot more predictible and easier to use. The only downside of using dirtmap is that you have to comp everything in layers but I acually prefer to do it that way because It gives me a lot more controll of the final image.

[demolisher]

hi guys thanks for the replies
thanks GGA for the detailed answers

"You are only opening a single file once and then spitting out all layers, right? If not, you do not understand what layer rendering is.
You mention episodical work. If you are dealing with a 30 min. tv series, truth is that you should be looking at keeping your render frames at 3 mins or less.
In that case, any sort of global illumination of any renderer will simply be too slow for you. You should stick to scanline rendering mainly and use raytracing also sporadically.
In a renderer like mental ray, you should also try to build your geometry with nurbs or meshes with subd approximation, and let the renderer handle the tesselation at render time (ie. not use things like polysmooth and similar, which will slow both the rendering and the opening of the scene file)."

well the files are so heavy i cant understand but mental ray is even calculating for hidden geometry i have no idea how so we delete the existing polygons and clean the geometry to reduce the render time
its like very heavy files we have 15 characters and very hi poly we got the meshes from our client we cant help it sorry

it takes ages for renders to start

well i have some more interesting doubts

have you guys ever tried the MI shaders

well were are the connections going to the shaders

1) if i connect a MI shader to the material it doesnt show a link
and when i say delete unused nodes the mi nodes are gone why is that any idea

2) has anyone tried 3d motion blur in metal ray it gives jagged edges i have no idea why

3) i have some problems with vector render aswell i get problems with animated characters when they move the render output gives me streak lines the option which creates the problem is "detail level preset" i make it HIGH and boom the mesh goes haywire

but in maya software its fine

the mesh is in poly and with smooth and the model is rigged and already animated
the image is attached

please guys i know you people may find the answer for these problems thanks guys for responding

thanks again gga

[LehaS]

I ll answer the 1st question ....you have to connect your MR shader into your SG by connecting .message attr of your shader to .miMaterialShader of your SG....That way it will be recognized by MAYA

Also....In order to use Volumetric MR effects, Photon Shaders and so on you need connect MR shaders to correspondent MR slots of SG...

[Karlfucious]

I personally would stay away from 3d motion blur in Mental Ray or any other renderer for that matter. Reason being is that it tends to be SLOW. If you need to use motion blur you are probably better off just using a specialty pass and Faking it in a compositing app. It may not be as accurate but it is verry close. Plus the moton blur pass renders in sec. per frame vs min or even hours per frame the normal way straight out the box. Im pretty shure If you do a search on the Maya rendering forum you should be able to find a great motion blur mental ray shader that does everything you need.

To answer your question about Vector rendering. The reason you would want to use the vector render is probably only if you need to export your vector imformation to Flash or something like that. It has the ability to export all of your contor lines and such for further work in Flash but if you are just going for the toon shading look you probably dont want to use it anyway. It is much less expensive to just build a toon shading system with ramps and or use the contor rendering system with mental ray.

If you plan on doing any kind of film work you should definatly look up some tutorials on rendering in layers. To my understanding it is pretty much the standard in the film indistry.
Out the box Maya's layer rendering system is kind of crappy and it really doesnt give you what you need. What you need to do is make multiple copies of your scene with all your specialty passes. Or if you are using Maya 7, It has a new feature that gives you the ability to do the same thing all in one scene which can be really handy.

One more piece of advice is to looking into command promp rendering. For one it saves a lot of memory that you can use to speed up your renders. Also since you are doing 30 min tv shows you will have a lot of hours devoted to just rendering so you probably will not be able to handle it your self so you will have to send it out to a render farm. In that case you will probably need to send them out a text file that sets up your batch job. Otherwise you will have to devote over a thousand hours or more per episode. I think I calculated 1440 hours of rendering time for 20 min of animation at 3min per frame with 24fps. My math may be a little off but you can see that It isnt really plausable to do without using a renderfarm where they can partition the render job across hundreds of computers.

Good luck,
Karl

[Lorecanth]

If you're doubting mental ray its just becasue you havn't explored it enough. The technical documentation in the maya help is dry, but mental ray does give you and extrordinary amount of power.

[gmask]

The only downside of using dirtmap is that you have to comp everything in layers

Not true .. you can connect the dirtmap node to your diffuse channel for example and get the same effect in one pass.

[danielkenobi]

Ok, about the radius that everybody is taking about I think that it sould not be meashure in Cm or inches I think it should be measured in units, because it doesnt matter the size of the escene,becaus is not necessary a scene to be real scale, I say this because there are other solutions that there are real world meashures, like Max radiosity.
Any way I think the best size of the radius for GI is 1/10 of the closest shot that you have, I mean the default radius of GI is that one 1/10 the size of your scene, but not considering that the escene can be really big or small touhg, compared to what the camera sees, that is where I am making the suposition. Altrouhg the reference of 3dmax show an example where the radius is preety big compare to the scene and they get good results with out FG. So what do you thik?

[revilo3D]

You are better off creating your own lights in the scene, area lights are extremely expensive to use.

A few suggestions. And a guess at my understanding of how "emit lights" works. I beleive it actually looks by default at a 256 x 256 resolution your HDR, and then creates a light source from each or these... eg - 65536 lights


By turning down the U and V sampling to like a much lower value will GREATLY speed up the render time.. however....


If you use HDR shop for your HDR images - then you can use the lightgen script for it to gerenate a MEL that will automatically create a dome rig style setup with lights,colours, intesites of your HDRi. You can then load into maya via HDRShopLightControl and have your scene based off your image and have complete control over the lighting it uses!...


An older but ingueous way of doing it that is still the winner today after all this time.. :)

http://www.highend3d.com/maya/downloads/mel_scripts/rendering/misc/2306.html

[floze]

An older but ingueous way of doing it that is still the winner today after all this time.. :)

http://www.highend3d.com/maya/downloads/mel_scripts/rendering/misc/2306.html
Well I havent used the one you mentioned yet, but I'd vote IBLTools for the winner:

http://www.thereisnoluck.com/plugins_iblTools.php

[Jozvex]

Well I havent used the one you mentioned yet, but I'd vote IBLTools for the winner

I second that vote! Do we hear a third??

[leftyfallat]

I agree that lightgen is a much better solution to using HDR lighting as opposed to IBL light emission.

It does basically the exact same thing as light emission but actually gives you lights that can be tweaked at your discretion.
I would say that the only downside to lightgen is that you are limited to a 128 X 64 image that the light information is calculated from. However, when light emission mode is used, render times are HORRRIBLE and if the results aren't what you want, you can't tweak the lights. You are then forced to tweak your HDR image rather than your lights...and then you have to sit.....and wait....I usually just use a combination of lightgen and final gather with the ibl sphere...crank up a light here, turn off a shadow there.

[revilo3D]

mmmm.. i think ill third that. HAd a look - tasty indeed :P

[floze]

I would say that the only downside to lightgen is that you are limited to a 128 X 64 image that the light information is calculated from.
I'm not sure if we're talking about the same program. The new one (it's called lightMapGen) is using the median cut algorithm, which is, compared to the other technique, incredibly fast. You dont need to resize your hdri anymore - in fact, the author suggests to leave everything at full resolution; the only thing you have to care about is the right panoramic format, which is longitude/latitude.

Also, there is a lightshader called sphericalLight. It basically does the same as the maya IBL light emission. There only seem to be problems when using maya standard shaders (which is kinda funny, because maya IBL causes problems with mr shaders..).

[leftyfallat]

Sounds like a pretty good program. We don't have that installed on are computers right now. We're still stuck with the old lightgen plugin which I here is hard to even find anymore. Is lightmapgen free?

[caulfield]

Hey Demolisher - I'm only using mental ray for a smeller job at the moment, and it's already got me scratching my head.

But I'm pretty surprised that you haven't seemd to have done any real RnD. Have you read all the docs? Books? Watched the DVDs?

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.