Jump to content

Tornado R&D


Fabiano Berlim

Recommended Posts

Just want to show you guys.

http://vimeo.com/33612394

http://vimeo.com/33612293

And v01: http://vimeo.com/33611981 - Different shape from the above.

Simulation time is around 20 hours in a i7 8 Core 12G of ram to 240 frames of 300x3 bounding volumes.

So you can imagine that is very hard to art direct. (I'm trying to solve this problem now.)

I've done a custom solver to only compute what is needed. It was taking more time to render in a regular smoke solver.

Any critiques and questions are appreciated!

Hope you guys like it.

Cheers!

  • Like 1
  • Downvote 1
Link to comment
Share on other sites

Hi there Fabiano.

Man I must say it´s been quit some time that I haven´t seen something so cool here at odforce.

It would be really cool if you could give some details.

Did you use pyro FX?

I tried some ideas like this an the sims are really heavy.

I see you have alot of patience.

Also how did you set up the volumes. I see that they are sticking really good to the sim and have no swim effect on the shader.

Well if it´s your secretes. I can understand

chhers good sim

mikey

Edited by mangi
  • Like 1
Link to comment
Share on other sites

Hi Fabiano,

they are wonderful!!! :)

If I have something to say, after watching all three versions (yeap, I was on your Vimeo channel), the first one is most interesting, because I like the dust forming at base, I like how it's animated and its dimension compared to the tornado, it's perfect for me and it should be more present, in a big scale, also in the other simulations.

I like also how it moves and bends, making it more realistic.

What I don't like is that it seems more "smoke-ish" than the other two, I mean, the version 3 seems to have more details, I like how version 3 appears.

Anyway, if I have to choose between version 2 and 3, I prefer the last one.

Why have you change the dimension from version 1 to version 2 and 3? Is there a particular reason?

How many hours did they take to render?

  • Like 1
Link to comment
Share on other sites

Hi there Fabiano.

Man I must say it´s been quit some time that I haven´t seen something so cool here at odforce.

WOW! Thanks! That's a badass comppliment! Thanks a LOT!

It would be really cool if you could give some details.

Is this pure dop forces or do you have particles controlling the tornado shape/motion?

I second mangi's post. If you could do so that will be really great.

Yes. It's totally dop. No particles at all.

But there's a problem with the normal volume forces with a normal solver. Smoke behave as it has mass. So it takes longer to start spinning and after it kicks in, centrifugal force pulls volumes out. That's very very hard to control. Since, the fast it goes, the centrifugal force is greater. You cannot turn the mass to 0. Since you don't have the property mass to volumes.

So to bypass it, I realized that I need to have a vel field that don't inherit the vel from the previous frame. You can do it using a 'gas field vop' and output vel without input vel.

That's not the most correct volume simulation ever. And if I turn down the velocity all of a sudden, the smoke would just freeze.

Inside the 'gas field vop' I have all the motion I need: A curve from sop that enter as 'point cloud open'. This curve has a direction attribute following it's own shape. With that you can just cross product this attribute with the position of the voxel to have a circular motion.

To be perfect: cross_product(normalized(curve.position - voxel.position), normalized(curve_shape_direction))

In top of that I just add a noise or several. Better to use at least one 4D noise (flow noise) to make it look more natural.

But if you just put noise on top, your tornado could just go bigger and bigger, not controllable at all. Also, if you move your line, the central point of your circular motion would move. And it would also make it bigger and bigger. To solve that I put a small force to push the volume to the center of the line. But that causes another problem. If your volume density crush into one place, it sinks. So you start to lose density. Not good! To override that I put another force slightly bigger to pull the density out from the center of the curve. That in the end is what maintain the diameter of the tornado. Knowing it, I put a spline ramp to control the diameter along the tornado.

I also made another spline ramp controller that makes it spin with the velocity I wanted along the tornado path.

In the end, since I don't need a real smoke sim, I made my own solver that just move the voxel based on it's velocity. I just use 5 or 6 nodes to make it. Very simple stuff. But it makes the simulation goes up to 5 times faster.

That's all!

Looks really nice, maybe it needs some more voxel detail, did you use upres sim for this?

No, i didn't, but I should've. I spent the whole year upresing! Kinda tired of it. That's why I went for the custom solver. Houdini 12 will save us all!

Hi Fabiano,

they are wonderful!!! :)

If I have something to say, after watching all three versions (yeap, I was on your Vimeo channel), the first one is most interesting, because I like the dust forming at base, I like how it's animated and its dimension compared to the tornado, it's perfect for me and it should be more present, in a big scale, also in the other simulations.

I like also how it moves and bends, making it more realistic.

What I don't like is that it seems more "smoke-ish" than the other two, I mean, the version 3 seems to have more details, I like how version 3 appears.

Anyway, if I have to choose between version 2 and 3, I prefer the last one.

Why have you change the dimension from version 1 to version 2 and 3? Is there a particular reason?

Totally agree! But for now it's a little bit hard to guess what's going to happen after 20 hours. Or maybe I need to improve my abilities of VFX Animator. I'm not sure how big guys can do it? Maybe patience.

I sent the first version to Miguel Salek. As a friend he told me that the tornado was great, but he didn't like the base. He also mentioned that it needed more sense of scale. I took it literally. For me it looks more dangerous then the first one. But the base sucks even more! :P

How many hours did they take to render?

I lot less then to simulate. Around 8 to 10 hours in the same machine. I cranked up the volume step size to 0.01 and my tornado is 7 units height. Also change the volume filter from gaussian to sync.

I can't share the hip file right now. I'm aiming to find a job outside of Brazil, and that will be one of the things in my reel. So for now I want it to be exclusive. Hope you guys understand that! I am planning to travel to London in February and advertise myself. If I succeed, I promise to up it in the Houdini Exchange.

But if you want to build it from scratch, I am happy to help you.

  • Like 1
  • Downvote 1
Link to comment
Share on other sites

Fabiano

Good luck on your Houdini Job .

There´s alot of production in the UK.

It woukld be really cool if you could direct us from scratch, Well I guess that if we have time.

I take it that you used the regular smoke container and not the Pyro ?

mangi

that sim is really worth the while,

But if you want to build it from scratch, I am happy to help you.

Edited by mangi
  • Like 1
  • Downvote 1
Link to comment
Share on other sites

Fabiano

Good luck on your Houdini Job .

There´s alot of production in the UK.

It woukld be really cool if you could direct us from scratch, Well I guess that if we have time.

I take it that you used the regular smoke container and not the Pyro ?

mangi

that sim is really worth the while,

Thanks!

Yes. If you want to build it. I am here to help. :D

I used a regular smoke object. The smoke object is the same for the pyro solver and for the regular smoke solver.

Cheers!

  • Like 1
  • Downvote 1
Link to comment
Share on other sites

it would be more controlable if you using particle to control tornado's motion. the particle is much faster than fluid for visualizing the motion of the tornado. when the motion and the speed is satisfy. send the particle directly to dop. add vortex force and gas dissipate microsolver to it. and pretty much done.

  • Like 1
  • Downvote 1
Link to comment
Share on other sites

it would be more controlable if you using particle to control tornado's motion. the particle is much faster than fluid for visualizing the motion of the tornado. when the motion and the speed is satisfy. send the particle directly to dop. add vortex force and gas dissipate microsolver to it. and pretty much done.

Has anyone had much luck with emitting fluid attributes from particles for tornados in Houdini? I've seen a bunch of real nice results from Maya fluids. And yeah, as you say, much more controllable

  • Like 1
  • Downvote 1
Link to comment
Share on other sites

Thanks!

Yes. If you want to build it. I am here to help. :D

I used a regular smoke object. The smoke object is the same for the pyro solver and for the regular smoke solver.

Cheers!

Hi again fabiano. cool and thank´s.

I did not really know how to dig into your technic.

Okay reading over your expanation, you mentioned:

"Inside the 'gas field vop' I have all the motion I need: A curve from sop that enter as 'point cloud open'. This curve has a direction attribute following it's own shape. With that you can just cross product this attribute with the position of the voxel to have a circular motion."

What I understand from this is you hooked up a gasfieldvop1 In the Dynamics.

Whould this be hooked up in the smokesolver in the second input ? Velocity Adjustment ?

Well If you could give some hints as we go on with out making anything really obvious.

that would be really assume:

Any explanations, tips , papers. etc..., for this discovory.

Okay Fabiano this may be fun to figure out this puzzle.

Drop in your hints when every you can.

Or papers you

cheers

mangi

  • Like 1
Link to comment
Share on other sites

Has anyone had much luck with emitting fluid attributes from particles for tornados in Houdini? I've seen a bunch of real nice results from Maya fluids. And yeah, as you say, much more controllable

Yeah im playing with right this now - the vortex sop works great - and sourcing fuel from particle animation to burn in pyro sim

But i find that adding the turbulent detail ends up with too much smoke tendril effect - or too much dissipation ebbing away from source - im still playing but this technique here is really sweet - as all the density stays contained and im guessing probably faster to sim - im going to try and recreate - excellent effect.

Link to comment
Share on other sites

Sorry for the late reply, I've been working a lot.

Okay reading over your expanation, you mentioned:

"Inside the 'gas field vop' I have all the motion I need: A curve from sop that enter as 'point cloud open'. This curve has a direction attribute following it's own shape. With that you can just cross product this attribute with the position of the voxel to have a circular motion."

What I understand from this is you hooked up a gasfieldvop1 In the Dynamics.

Whould this be hooked up in the smokesolver in the second input ? Velocity Adjustment ?

Actually you can plug anywhere: Velocity Adjustment, Advection Stage, Post Solve or you can either plug anywhere after the solver.

Well If you could give some hints as we go on with out making anything really obvious.

that would be really assume:

Any explanations, tips , papers. etc..., for this discovory.

Hard to say. I didn't try to find any paper about the subject. I only made a research about how tornadoes works and looks. That means: Wikipedia, YouTube and Google.

The math and Houdini stuff behind it, was just what I already knew: Basic geometry operation, point clouds, circular Motion, noise, dops, vex, etc.

I can give you some hints: The point sop node have an Edge Force paramenter. If you turn that on, it will create a edge_dir attribute. This attribute will give you the direction of the curve.

Check this file:smoke_spin_FAB.hip

It'll help you to start.

Cheers!

Link to comment
Share on other sites

It'll help you to start.

Cheers!

Hey there fabiano, Thanks for this starting point.

It´s really cool.

I take it that the Tornado shape is all done in the geometry node ?

Also the effect of the shading . It´s not the regular smoke billow shader ?

Can would give some details on how you approached this.

well if get some time.

Or maybe after xmas

cheers and happy xmas from madrid

mangi

Edited by mangi
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...