motion blur in post

#13

another intersting thing of note I must ask. I looked at the help section of blur2d.exe and there is no mention of a “-v” flag. I always thought that i could use motion vectors to blur other image sequences, but could never figure out how to do it. What is the -v flag?

#14

You have to remember… this is a blatant hack to get around the ridiculous render time of MR motion blur. You can’t expect it to be 100% accurately perfect. However, it is an extremely fast solution… With results that have been used in film. They look very very good. Yes, as long as you process an alpha… Your blur should be quite close, if not… the same… as to what MR punches out for you… When you’re in comp with it hooked up to RSMB 3.0 Vectors… It even processes the filter fast (depending on how crazy your comp is, I’m just going off of layer/source/result) … Either way… its saved hundreds of hours of waiting for me and other people here at work. As for blurring shadows? … You may be S.O.L. on that front unless you can think of a way to get it to spit out a SHADOW for the blur instead of a surface/etc… if you get what i mean …

oh, and it will not respect displacement mapping …

az

#15

I cant seem to get this to work in Digital Fusion using the ReelSmart Motion Blur v3 plugin. Im using Maya 6.5 and trying to use use the lm2DMV_v2 Mental Ray shader.

It renders fine within Maya, although it wont seem to render an alpha image (not sure if it is even needed), but the RGB seems to look ok. I then take it into fusion and try to work with the vector blurs and nothing happens. BUT when I put an alpha in there (channel boolean tha alpha from the beaty pass) then the blur works, but the blurs are wrong (i assume because of the worng alpha)

Has anyone had any luck with the shader and using it to blur in post, with or without Reelsmart Motion Blur?

Interesting thread guys, anything to speed up MR helps!

Thanks.

#16

read above to my first post, it explains how to force the rendered to generate an alpha. and yes, once again… an alpha is required for results… and using a track matte for an alpha doesn’t seem to do the trick… i’ve tried this myself in the past…

yes, I use it almost every day flawlessly.

az

#17

ooop!

sorry azshall (i hate people who dont read whole posts, looks like i’ve become one of em )

THANKS!

#18

Also, if 2D blur doesn’t give you what you need, have you tried the various motion blur options in mental ray?
Do you know how all this is linked to samples?

Have you adjusted time contrast, which is really just a way to specify number of time samples per sample?

Have you tried setting time contrast to all '0’s to use a faster motion blur, which depends on a fixed sample pattern, ie min samples = max samples? If your scene is filled with moving objects, this should make a big difference.

And if you don’t need raytracing, using the rasterizer (ex- rapid motion) is another faster option.

#19

Works a treat now, thanks!

One other quick question. Is using motion vectors a more accurate way to simulate motion blur in post over using Reel Smarts actual motion blur algorithms through comparing successive images?

Just wondering.

#20

what should be done with the general sampling and contrast values (should I use the same as I did for my regular rendered frames or is it ok to go with less quality here…)?

#21

the problem with a fast motion blur (min samples = max samples) is that you dont have to much control over quality and its not good.consider i have a production that i need max samples set to 2 (in must ot the cases)then i should use a min sample as 2, and its expensive for render time. and rasterizer (ex- rapid motion) is a joke i cant imagine to use a MR withut raytrace (GI-FG). so the only solution is in post.lm2DMV_v2 for maya is realy cool the only limitations are 1.dont support shadow 2.dont support reflection and refraction, i think the best way to work with lm2DMV_v2 is to have as much as possible objects in a seprate layers and use lm2DMV_v2 per object and then in a compose soft use Reel Smarts for each layer(aranging layer in a depth in realy important). with his technique you can have a lots of objects that have a extream motion blur and are over each other and not conflict with each other (remember the beauty layers for compose must be separate too).the same technique should use for a correct dof effect. if you use only on single zchanel for the entire scense you will always have a glow effect around the edges. anyway i realy need a solution for a shadow motion blur in post. anyone know?

#22

The -v flag tells blur2d to take the motion vectors from the secon file not the first one.

#23

wow - so many detailed replies over (euroean)night.
special thnx to azshall for your walk through.
it is a bit of a tricky beast, but also here it seems to work, now.
Ireally like the images that come out using it on a character.
like a thermal image. planing to do my whole short in this style :smiley:

#24

What do you mean by this? Rapid motion doesn’t stop you from using raytracing in any way. The rasterizer is just a highly efficient eye ray calculator which does motion blur in a different way.

It doesn’t stop you from doing anything else that mental ray can do, just performs eye ray calculation (and by extension also motion blur) in a different way.

/Z

#25

yeah the motion vector thingy works reely good, i tryed connecting the shader to the shadow slot of the shading groupe but don’t know if this is correct, and the aplha wouldnt work inn this image. anyone else got an idea on how to do this?

btw i can’t get volume effects to render if i turn on rapid scanline methode for faster motion blure

#26

thanks [Master Zap]. but is rasterizer for a high quality production? its geting slow with FG or iam wrong?
[Osaires] if its work with shadow (that iam not sure) i will be :bounce:

#27

The rasterizer is for high quality production. It uses a different method for rendering the objects seen by the initial straight eye rays.

As Zap said, you can combine it with raytracing as well, if you wish.
Its just that the added raytracing will add to the render time, as it must then create a BSP structure, etc. FG rays need the same structures to be set up.

With a scene requiring raytracing, I would like to see more people try the option to set time contrast to 0 with fixed samples.

On the other hand, if the scene has just one moving character, the 2D post blur seems like a straight forward way to go.

#28

[Osaires] If you’re just using the rasterizer because of motion blur, try the time contrast 0 option. This will not change the behavior of the volume effects.

If you set min/max samples to 1, then the scene averages a density of 4 samples per pixel.
With min/max samples set to 2, then its 16 samples per pixel.

The motion blur is accomplished because each of those samples occurs at a different time in the frame.

#29

Motion blurring of shadows in its own pass using the useBackground shader isn’t expensive to render though. So I dont find much of a need to do it in post.

#30

Hey azshall, thanks a lot, now the shader is rendering properly. But I didnt get the chance that to use the MV. I doubt blur2d.exe would recognize them.

For Motion Bluring the shadows I really dont see the problem in that. What really bug is the moving object wich is the subject in the animation. Shadows come in the next and is seen as something additional. Unless you’re subject is the shadow itself.
I’ve seen in several 3D feature films non motion blured shadows and it didnt affect the quality at all.

#31

Sorry to warm this thread up again, but I am trying to use the lm2DMV mi shader for my motion blurs and all I get is a 0 pixel displacment (olive green) picture. Does anyone know as to why this could be? The shader seems installed properly, i have MB switched to EXACT, camera movement, transform movement and deformation in my scene and still doesent the lm2DMV shader recognise any movement. When I render a beauty pass with MB everything works fine… :shrug:

Would be great if anyone could help with this…
thanks,

Jens

#32

Did you read through the thread from the beginning?

http://www.cgtalk.com/showpost.php?p=2243447&postcount=10

az