View Full Version : LW9.2 and the new motionblur...

07-12-2007, 01:59 PM

After some sweaty rendering with MR (my god, that renderer can do wonders if you spend hours and hours of tweaking) I've found out that the motionblur doesn't look great, and certanly, it does NOT render fast. There are two types of Motionblur in MR for Maya; Linear and Exact. The Linear renders alot faster than Exact (thus, it doesn't look half as good). But..the funny thing is... a frame that renders in 20 sec without motionblur can take 3 hours to render; using Linear motionblur!!! (I'm not kidding here). You REALLY have to tweak to get it down, but...insane stuff! So, you can imagine that I don't have any interrest at all in trying the Exact setting... I mean... if Exact is slower than Linear... Oh My God... no wonder they use Renderman in 4k res movies/vfx.... :) :) :)

So... my question is... where would LW:s new motionblur be placed compared to the two giants, MR and Renderman? Is LWs motionblur more to MR's Linear or Exact Mode? Or is it even more close to Renderman?


07-12-2007, 03:10 PM
Probable somewhere in the middle of all three.

07-12-2007, 03:11 PM
That would mean MR:s Exact Mode in other words? :)

07-12-2007, 03:59 PM
It's fairly hard to compare them. Mental ray has a number of methods each with there set of draw backs. But mostly it all comes down to how each app handles it's sampling. In terms of LW and Mental Ray they have to deal with Raytracing. Where as renderman handles all of it's raytracing on the shader.

All in all i think LW comes closet to Renderman for AA in terms of features.

If it helps here is a post on the XSI mailing list about Motion blur in Mental Ray.

There was a good summary of the options a while back posted by Harry
Bardak, I simply copy&paste:

"There are different way to tackle motion blur in XSI :

Option 1 :
You don't want to use Rasterizer, you want to use standard Motion Blur.

Rendertimes ( and also Quality of the MB ) Will depend mainly on the
motion threshold that you got in the motion blur tab.
By default it is set to 0.2 0.2 0.2 0.2 which means that 5 sub frames
will be calculated and merged. If you put 0.1 0.1 0.1 0.1 10 frames is
You got more info on this at www.lamrug.org.

So basically if you put 64 samples on your AO shader you will have 64
x 5 samples calculated which is way too high. So you can divide your
sample number per 5.
Activate jitter. You will get the equivalent quality as when you don't
use MB and also fast render.

Option 2 :
Same as above but you control of the amount of sample according of
motion vectors.
Guillaume Laforge did an article about this on xsi blog. It's one way to do it.

Option 3 :
Use rasterizer. To push too high sampling rate. ( 2 - 4 ) otherwise
rasterizer will struggle with raytracing ( it s getting faster and
faster with every version of MR by the way )

Option 4 :
Activate Fast motion Motion blur. This is a special mode that don't do
any adaptative sampling. Put your min and max to same value and motion
threshold to 0 0 0 0.
This strategy can be useful when you image is filled by a lot raytrace
sample like Occ or glossy reflection. MR in this mode will not spend
time to adapt the sample to what is needed or not.

Option 5 :
Camera projection technique. It's a 2 pass approach. First you
calculate you image without Motion blur then you camera project the
result with a constant shader. The motion blur will be very fast in
this case.

Option 6 :
Post procces approach by outputing motion vectors. You need RealSmart
motion blur plugin for Afx, Fusion or shake. Or you can subscribe in
the beta shader forum on lamrug, and download the mi production
You will have to generate the corresponding Spdl. This shader is an
output shader that add motion blur after you render.