Mental Ray with motion blur

[Saturn]

a little trick to speed up motion blur 3D with MR.

Render your pass with all option you want raytrace, GI, FG whatever, but without MB. When you are happy with it.

Setup a new scene and do a camera projection of your previous pass to your object. Object material must be self illum, Turn off all light, all shadows, turn off all option, raytrace GI, FG etc...

Tweak your Motion blur setting, and render.
MR will just calculte MB, without computing lightning, raytracing etc ...

It's very fast, i use it every day but on xsi.

[a23]

You can also try the 'scanline rapid' option which does shading before hiding with some caching of the samples.
this speed's up motion blur quite a bit. although not being a 100% accurate you will probably not notice it. give it a try.

at least it work's great in xsi, so max should work as well. check your options and read the MR docs.

[Tora_2097]

Thanks for your comments.
I did some research to find out more about MR's motionblur and in this (http://www.zaon.com/company/articles/3d_rendering.php) you can find that the motion blur speed in MR has been claimed as very slow. So It seems that this is one of MR's weak points.
I made a testrender yesterday of my Aston Martin (see it here (http://www.cgtalk.com/showthread.php?s=&threadid=126164&highlight=aston+martin) driving along very fast. The render without motion blur took 1:34 min at 640*480 and with motion blur enabled with 0.8 shuttersize and 5 mtion segments it took 59:47 min... The car itself has 330000 polys all included. As I am only willing to pay max. 15 min/ frame I guess I really have to find a trick here.
@saturn: I am not too sure if I understood you correctly. Could you please explain it a little more detailed? I tried it out yesterday, it worked, but where is my timesaving? Motion blur is even slow when I disabled everything ( raytracing, shadows etc..)
High polycounts let rendertimes explode in a unacceptable manner.
@a23: I wil try that out, thank you.

[a23]

Well, with Saturn's method, which i've used in the past, motion blur should be very fast. of course you have to be sure to disable raytracing during this pass. i'm sure he will explain it to you in more detail.

The technique 'mi' have introduced in MR 3.2 they call 'scanline -rapid', typically has a slowdown of about 100% when using motion blur. which means twice the rendertime. You should also take a look at your ray depths and don't have them too deep.

but it pretty much depends on the implementation of MR in Max. Since even in XSI some important options to this feature are not in the UI.

Good Luck.

a23

[Tora_2097]

thanks a23, therapid scanline option decreased render times a lot but the results look unsatisfying to me. though it might be very good for testrenderings and finding the right settings.
With "normal" scanline and motion blur rendertimes increases- compared the non blurred image- around 40 to 80 times. disabling raytracing, shaodws, reflection, etc still needs around 15- 20 times of the non blurred image... :shrug:

The mental ray implementation in 3ds max was considered as " very good, only one step behinds XSI's" which was considered the best.

[MasterZap]

If you by "normal" scanline mean the MAX default scanline renderer, motion blur should increase your render time by nearly nothing IF YOU USE IMAGE MOTION BLUR, which you should. No sane person use object or camera motion blur (which IMHO are holdovers from days of yore).

Just a tip ;)

/Z

[Tora_2097]

As I posted in the beginning I would like to use image blur for most of the parts in this animation but as want to render the scene with Mental ray this task becomes impossible because Mental Ray does not support image motion blur. MR does only support real 3D motion blur because some of max's G-buffer channels are not compatible with MR.
With "normal" scanline I did not meant the max standard scanline but one of the two MR rendering options callled "normal" and "rapid" scanline.

[lazzhar]

Why not try to fake MB in post using Compositing Programs. There is also some cool plugin cannot remember its name right now but it produces very interesting results IMHO.

[Tora_2097]

Thanks a lot guys, but I solved all my problems in moving from Mental ray as a renderer to Final render stage 1 which I got yesterday.
It gives me better results faster with superior image quality.
Now I can combine 2d and 3d motion blur in one pass and I have very accurate controls over all needed aspects. The entire environment uses image motion blur whereas the car and the tires uses real 3d motion blur. Final render 3d motion blur works 7-10 times faster under the same conditions compared to Mental ray!

[Saturn]

I don't why polycount slowdown with my method
I had a scene of 1.5 Millions poly to render in 2K
It take me less than 5 min per frame with standard scanline and at most 2 min per frame in rapid motion. ( On a P4 2.6ghz with 1gb ram, fast if you as me)

Can you describe me how did you setup your scene ? I mean where do plug your image in wich shader ? I suspect that you used wrong parameter that oversample you pic.

Also, what AA setting do you use ? wich Time constrast ?
In rapidmotion what's your max value of AA ? your shading rate ?
Anyways if you find solution with FinalRender and happy with that go there !

Last thing : Max implementation of MR is far away from Xsi's one.

[MasterZap]

Question: Why on earth would you use 5 motion blur segments!? Thats just mega mojo overkill. 99.9% of the time 1 segment does just fine.

/Z

[Saturn]

i guess it s for the wheel ...

[Tora_2097]

Thanks a lot for all the replys. :)
@saturn: Yes, I might have set up something wrong. Although I will stick with Finalrender in this situation I would like to learn more about your method here.
But I have the feeling that MR in XSI and MR in 3ds max may differ a lot. For example 5 motionsegments (yes, I use them for the wheels) in max have only a small effect on the rendertimes, whereas high polycounts kills my PC.
Maybe you can explain it again to me.
I use an Athlon XP 2600+ with 1024MB RAM so it should be around as fast as yours.
I usually use AA filters at around 1/4min to a max rate of 4-16. I use the Blur AA filter in this case. After you method I should render my scene at first with raytracing, shadows etc. enabled but without motion blur correct?
Then I do not understand this sentence of you "Setup a new scene and do a camera projection of your previous pass to your object" What Do I have to do exactly? Most of the times it is not only one object (the car) that has to be blurred but mostly the entire environment (1000+ objects)... How do I achieve that? And how to combine these two passes?
I am sorry if my questions sound stupid. :blush:
Thanks in advance for any help!

[Saturn]

First part is correct. render your scene as usual but without motion blur.

Second part is easy also. Take the same max file. Create a new material who will be self illuminated ( my english is correct ? ), apply you first render ( the sequence ) on this material. Then all object in your scene should this material
Now you do a thing, i don't know how to call that with max. Basicly you have to use your camera to project this first render on your scene ( as know as camera mapping )

[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.