Sand blowing over sand dunes

[xtremeDave]

Hi Guys,

 

 

 

like this halo trailer , have any of you tried this kind on sand blowing over the sand dunes Effect ? and  if so would be good to know the most efficient way to get this done,

 

and also , if any houdini Fx guys are looking for work PM me ,  mainly sand effects !

 

Thanks,

Dave

[Skybar]

Looks more like a smoke sim rather than actual sand. I dont imagine it being too difficult to achieve.

[eetu]

That trailer was actually done in Houdini, by the fine folks at Axis Animation. Maybe the guilty parties will step forward

If I remember correctly that indeed was 'just' a smoke sim, but getting things to look just right never tends to be real easy..

[chrobak]

I don't imagine it being too difficult to achieve. I would start by using pyro with wind force and messing with the settings and mix it with particle in dops interacting with the static volume collision geometry of the dune. Use two separate sims with similar wind settings. 

[NSugleris]

After seeing this over the weekend I decided to do a small test with vops particles and fluids  

 

[JuriBryan]

well I think it is like eetu said a "simple" idea, but to get it to work and to take it from a tech demo to an actual production is very hard and takes a lot of time.
In the end it is always like testing water stuff with a cube falling in a flip, everything can work there.
but take that setup into a shot and all hell breaks loose.
 

[pclaes]

Mostly I think it is about data management and getting the scale right.

It is about layering different noise fields that are offset in time and animating them over the surface. So you have different frequencies of velocities crawling over the surface, advecting your fluids.

Then I would source them from points and partition them based of the area that you need to cover.

For really close up shots I would advect particles and mix that in with the volumes. 

 

The way ILM gets it right is by wedging sims with different parameters and picking the one that looks best, wedge again on that version and refine.

You can wedge on a single cluster before running all your clusters.

[xtremeDave]

Hi Guys,

 

Good start test Nico, 

 

As some of you have said , getting production quality results is what we are after !

 

So do you guys think thin smoke , particle advected by flip fluid the best and fast way to approach this and other sand effects like sand piling up ?

 

 

 

 

[NSugleris]

Thanks, its not what you wanted but after seeing the post I wanted to try out dissolving a mesh into sand. The setup I used had the fluid being advected by the sand particles and vopsops for the dissolve.  

Edited June 10, 2014 by NSugleris

[JuriBryan]

I personally would go with a decent res smoke sim.
do all the look deving of the motion on that/those sims.
Then advect a shit ton of particles with those sims.
render them with a constanst shader that catches shadows and a very small point size.
 

[xtremeDave]

Thanks Nico, Juri , eetu and Peter, great suggestions ! 

 

As we need this to be fast as well , for sand effects , like in this video 

 

 

sandman formation , we have done many tests , but using flip solver 

A : its slow 

B: its more like wet sand 

 

and yes we have looked at the walking melting sand girl from a while ago , the issue with that was , it was very slow to calculate !

 

is there a better approach of sand formation / piling up effect / sand blowing away off a character (Nicos test was good start for the last part),

as in this video , and hip files would be good to analyse the best way forward,

 

 

along the same topic, id like to let you guys know, we have received many emails from Fx td`s ,

we are replying to the ones we think are suitable , that does not mean the others are not good , its just we need more of a sand solution guys !

 

Best ,

Dave 

[xtremeDave]

Nico , if you could post the hip file we could see what works and what could be improved,

thanks once again for all your feedbacks ! 

 

Best,

Dave

[JuriBryan]

 I would go with pyro sims... all the way, at least for mid/background elements.
and then layer a sim or packed rbds in for the forground.
that way you get the mass and all that where you need it and the motion where you feel/see it.
 

[MENOZ]

Hello, I did the blowing sand setup for the trailer, I can't explain in detail how was it done, but from houdini side it was pyro + particle advection, many different passes.

There was also a pass, quite heavy in terms of sim, of turbulent-swirling -ground level sand.

 

A triky part I would pay attention to is the turbulence created at the edge of the dune wich is probably sucking in all the sand. I would use some air fields or collision modelling to control.

 

There were lots of other artists that created this piece and lots of work was done to bring the final images. Each shot had its own challenges, it wasn't certainly a one-go solution.

 

anyway, I would use real life references, not cg.

 

hope this helps a bit! other users here at od force already told you all you need to know to get the desired effect, so not much to add really.

[NSugleris]

Hello, I did the blowing sand setup for the trailer, I can't explain in detail how was it done, but from houdini side it was pyro + particle advection, many different passes.

There was also a pass, quite heavy in terms of sim, of turbulent-swirling -ground level sand.

 

A triky part I would pay attention to is the turbulence created at the edge of the dune wich is probably sucking in all the sand. I would use some air fields or collision modelling to control.

 

There were lots of other artists that created this piece and lots of work was done to bring the final images. Each shot had its own challenges, it wasn't certainly a one-go solution.

 

anyway, I would use real life references, not cg.

 

hope this helps a bit! other users here at od force already told you all you need to know to get the desired effect, so not much to add really.

 

Awesome!  I was really waiting for the artist to come out of the dark

 

I'll submit a hip when I get home even though it's definitely nothing different than whats been discussed and is already well known.   

[tricecold]

I usually approach these kind of shots with the question, How much I can divide this sim into, 10 containers 20 containers, Then I usually setup a easy to change and adjust workflow, with ropnetworks, so when I am back in the morning I have a version to watch.

 

In some those shots, visibility wise we don`t see the source but there is always some sand rolling, 

I would make a small area test with Pyro with negative buoyancy and gas wind then I would advect the particles like many here adviced. 

 

But the render of the particles can be tricky,

I would try to avoid the billowy smoke accumulations.

 

After I get this test working, take your landscape into the scene and divide it to the sections, by camera distance to optimize, maybe 5 containers in front , 3 containers at the mid and 1 container at the back etc etc, whatever fits to your situation.

 

Then make a NULL with some parameters to control your 10+ dop networks, like Voxel size, wind direction, noise amplitude, POP source count etc, It will save you a lot of time when the system is working for iterations.

 

If you have a Hqueue Farm, plug in a wedge node then, chek back the next day.

 

After you all caches are ready, I suggest =or you to use Delayed Load Shaders for where possible. it will make the lighting tests faster.

 

Thats how I did similar tasks in the past.

[xtremeDave]

Thanks Guys, great info ! , please do send us your cv / reel , we are hiring !

 

, 

[rayman]

If you are in situation where you need stacking piles of sand with a lot of collisions between them, you can try different approach.

Recently I did some tests using bullet as a base instead of flip/pyro and the results are great - 10-20M particles controlled by lowres rbd sim and you have great control of initial sim and the speed is nice(probably it can be faster using bullet sop).

What you need for this setup is simple bullet simulations with a couple of thousands spheres or cubes. Simulate and cache only the points without any geo. Then create upres part of the solver using sop solver and pointwrangle as a base inside dops. Create milions of particles matching closely the initial positions of original ones. On the first frame get closest n drivers and store their ids as array attribute per point. Then on each frame get original positions and create some average position value. Set it as v for the pop solver or directly as a P (I prefer pop solver as it has built collisions and etc., but you can write your on version). Now the behaiour is really nice, and what you have to do is to check for drivers detach conditions. For example when you have 10 drivers per point and some of them are going in the same direction and the others-not, you can detach bad ones, removing their id from point attribute, and if you want you can check for new closest replacement.On the same time you can search for closest drivers and get some part of their velocities. The attach/detach possibilities are endless, and with some parameters and ui you can achieve many different looks form the same base sim. And its going to be really fast(for example 2 hours upres for 20 M particles/350 frames). I`ll post some examples when I go home.

 

Edited June 22, 2014 by rayman

[xtremeDave]

Nice one Pavel, very nice approach ,

 

when you say , (the same time you can search for closest drivers and get some part of their velocities) , do you mean point cloud approach ? , 

yeah an example would be great ,

as we are looking for the most efficient way to handle this, and we have got some good info from previous posts from the other users,

 

Thanks Guys !

 

 

[xtremeDave]

Hi Pavel,

Those images look great, if you could post a hip file , it would be very useful to study this method

dave