3
$\begingroup$

I'm looking into how to achieve forces in Blender particles and fluids driven by motion vector image data. Think "optical flow". This is possible in Nuke, Houdini, and other packages. It appears like the functionality may be there in Blender but I haven't been successful in getting it to work correctly. I'm wondering if some of you out there have had success. Thank you kindly for your help.

$\endgroup$
6
  • $\begingroup$ Could you include some reference in your question? $\endgroup$
    – cgslav
    Commented Nov 27, 2017 at 22:20
  • $\begingroup$ Well Houdini added this in recent 16.5 version. You can trace movement from video (optical features), but particle and fluid side of things is very weak. You will need Animation nodes addon and some custom scripting to get anything done, then also mesh those particles into fluid. You are talking about this: vimeo.com/208144537 are you. Maybe in couple of years (5+) this will be easier to do in Blender. $\endgroup$ Commented Nov 27, 2017 at 22:23
  • $\begingroup$ That is the process yes. Houdini can do this, correct. I'm looking for the approximate solution as it may exist in the blender workflow however. $\endgroup$
    – VFXTobias
    Commented Nov 28, 2017 at 0:30
  • $\begingroup$ p.s. I already have optical flow footage that I can create myself without a problem. Just trying to figure out how to make use of it in Blender for particle and fluids. $\endgroup$
    – VFXTobias
    Commented Nov 28, 2017 at 16:52
  • $\begingroup$ It can be done using Animation Nodes addon. I don't know if it can be done using the fluid simulator or any other part of blender. The process in Animation Nodes will be super complicated, at least for me, so I don't think it can be written in answer here, I am studying this exact topic now, so I may write about and link the article to you or write a smaller answer here. $\endgroup$
    – Omar Emara
    Commented Nov 29, 2017 at 12:18

1 Answer 1

-1
$\begingroup$

hmm just brainstorming. Convert motion vectors to points use those points (make sure to keep points to a limited number) to act as effectors in a fluid sim. So you have to probably write a script that converts pixel location x,y,z data into points. There is software out there but that's the extent of my knowledge on this.

I use this with the depth data from structure scanner to make a per frame point cloud. It might work with standard motion vectors. Its a start.

https://github.com/arthare/stac_sample_decoder

$\endgroup$
1
  • $\begingroup$ Does this actually work? Have you tested it or are you just hypothesizing? $\endgroup$ Commented Dec 22, 2017 at 18:47

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .