I got stuck on a situation where I have a nice grow effect using solver and morph VDB but I don't know how to generate a velocity field out of that for motion blur, any tips?

I mean how would be a good way to produce a vector field for motion blur from something like this?

https://www.youtube.com/watch?v=nDTkjbEwdOM [www.youtube.com]

or this:

https://www.youtube.com/watch?v=GnSdgyhlsWE [www.youtube.com]

Same. I have an object being revealed in a vdb and need motion blur. Guessing its the changing point count causing it, but could really use a solution.

Can you use a gradient of the field to sample as velocity? Or basically whatever velocity field you're using to advect the volume...

I think this will blur everything everywhere, this issue is a tricky one.

mzigaib
or this:

https://www.youtube.com/watch?v=GnSdgyhlsWE [www.youtube.com]

In this one he is literally using the gradient of a density field to generate a velocity field for advection. It seems coherent to me that such velocity field would be appropriate for motion blur calculations.

rafaels
Can you use a gradient of the field to sample as velocity? Or basically whatever velocity field you're using to advect the volume...

The object is being revealed with a box moving over it and SDF Intersect. Im also using MOPs to replicate a c4d cloner setup to add to the edge of the volume. Those are being scaled up as the box moves to reveal the object, but not sure if there's a way to transfer that motion/velocity onto the final vdb.

rafaels

mzigaib
or this:

https://www.youtube.com/watch?v=GnSdgyhlsWE [www.youtube.com]

In this one he is literally using the gradient of a density field to generate a velocity field for advection. It seems coherent to me that such velocity field would be appropriate for motion blur calculations.

It is hard to tell since I can not see any final render to analyse it but it seems to me it is going to fall in to the situation that I was talking about before where you have velocity everywhere you should have velocity only where it is moving and also there is the matter of the intensity of the velocities where you don't have any way to calculate where is faster or the actual acceleration since this is just coming from the information of SDFs gradient.

If I don't have a way to say and show to a sup exactly and clearly how motion blur is working when I send something like this I would rather not use it and here I don't think I can do that.

I see your point. I tried something quick and dirty and it'll depend a lot on how fast you'd be advecting your fields... but this worked for my hypothetical case:

• Mesh the current frame and the previous frame (with a timeshift);
  
• Subtract the current point position to the closest position on the mesh of the previous frame (e.g. using xyzdist in a vopnet);
  
• Then you can either clamp the velocity values based on the minimum threshold and smooth out all the velocity vectors to prevent any jumping around;
  

I can't post a file now cause my work machine is blocked from the internet, but I'll do it later on.
Again, I don't think it'll work for all cases especially if your growth pattern is very abrupt.

It is an idea, it could work to a certain point.

The thing to work would necessary need to know what part is growing and for what direction that is the tricky part.