[DOC-6765] motion blur on vdbs not working?

[LeoYfver]

Dont know if im doing something wrong here but cant get motionblur to work on vdbs

Regards Alexis

https://drive.google.com/open?id=1G6...u66IoeqrMuumnm

[Gosho.Genchev]

Hey, Alexis,

you can find the updated file here: https://drive.google.com/file/d/1vDS...ew?usp=sharing

The vel field should be split up into it's component fields using a VDB Vector Split - it's a limitation we can't work around at the moment.

Two more things:
1. If you notice clipping of the motion blur, use a VDB Activate in Expand mode - the blur is calculated only as far as the bounds of the container.
2. You may want to check out the "Backward Trace" motion blur method on the Rendering tab of the shader - the Phoenix FD dev team implemented it specifically for people asking for behavior similar to Mantra. More info here:
2.1. Thread about implementing this: https://forums.chaosgroup.com/forum/...ur-issue-w-vdb
2.2.The difference between the two methods (I'm quoting the relevant post) can be found here: https://forums.chaosgroup.com/forum/chaos-common/chaos-common-public/1043243-motion-blur-does-not-work-properly-with-vdb-volume-from-houdini

V-Ray calculates volumetric motion blur in one of two ways, both of which differ from Mantra:
1. By tracing the velocity for a specific voxel backwards and stretching the content for the newly found voxel at position (current voxel + velocity*-1). This is the closest to what Mantra seems to be doing, which is directly stretching the content for the currently shaded voxel.
2. The default option expects grid velocity information to be present around the content (e.g. density) of the blurred voxel. This works really well for simulations where the entire velocity grid's bounding box contains velocity information but may give you trouble if you've generated a VDB with a random velocity field right around the edges of the density. In such cases, the density will be stretched as far as the velocity vectors go, and will abruptly end there.

I keep thinking I've added this stuff in the documentation then realize it's not there I'm adding a ticket to update this.

Hope that helps!

[LeoYfver]

Thanks for the scene and detailed information once again!

[LeoYfver]

Bumped into a new problem. when shutter_offset is set to 0 i got no motionblur on my sim. An example scene with a simple fire sim

https://drive.google.com/file/d/1ZbK...ew?usp=sharing

[LeoYfver]

this one is problematic for us. Is it something that can get fixed?

Regards Alexis

[Gosho.Genchev]

Hey, Alexis,

this will likely take some time to be taken care of - the dev team that works on volumetric rendering issues have their hands full at the moment. I'll ping you next week with more info.

Cheers!

[LeoYfver]

ok, thanks for the update!

Regards Alexis

[LeoYfver]

Gosho.Genchev actually i think it works but the amount of motionblur change so much i looks like there aint any motionblur.will do som examples if I find the time

[Gosho.Genchev]

Yup, that is indeed the case. Something is not entirely correct there when the center is at 0 so a developer will have to take a look. I tried figuring out myself what exactly is happening but could not find a reasonable solution.

Please do not spend time on this - you guys are busy enough as is. I'll sync with the devs as soon as they have some spare time and we'll figure something out. I'll keep you updated.

Cheers!

[LeoYfver]

haha, allright sound good !

[Dubbie99]

Hi, has this issue ever been fixed?

We have had to give up on V-Ray for volume rendering on our current project. We just cannot get MB to render correctly.
100% of what we render is at -90 shutter offset so that it matches matchmoved cameras.

Mantra renders MB exactly like you would expect. It also allows undersampled vel VDBs and you don't have to expand the VDB to cover everywhere the motionblurred density will cover.

I would love to get an update on this.

[Morbid Angel]

We have the same issue with alembic caches of the meshed fluids, since those have changing topology they have to rely on velocity and the velocity does not work correctly. The proper setup to match the matchmove geometry and camera is to have shutter of 0.5 (180 deg shutter) and interval center of 0 not 0.5. When geometry has constant vertex count it can be sampled between frames and renders correctly with interval set to 0. But velocity renders wrong if sampled anywhere besides interval center of 0.5.

[Dubbie99]

yes - this issue also forces us to use mantra for vel based MB. It's really frustrating sometimes trying to use VRay for FX work.

[Morbid Angel]

We were diving into this today as it's preventing us from completing some of the shot work. We have a mix of cached alembics which use vertex motion blur and velocity based alembics which use vector blur inside the mesh due to changing topology. The velocity based alembic always renders wrong, except when interval center is set to 0.5. However I've loaded the exact same setup in maya and it works out of the box even when you set the interval center to 0. I am going to present this case to Chaos, hopefully in next few days and it should get resolved asap.