emFluid4 released

[Gith]

Hey origin, do You have trail or full version?

I'm wondering how You've achieved that detail. I'm currently at home and I was playing with it, but here, I only have trial version. Have You changed Cell Size maybe?

[origin]

I have full version. Although I don't know anything about this shader, yes I increased number of particles (about 100k in this examples). Cell size affects only quality of simulation.

[Mootzoid]

Hey guys,

just to avoid any misunderstanding:
when you use the BA_Volume shader to render the emFluid4 stuff then the amount of particles has absolutely no influence on the rendering !

What Holger and I did is the following:
We implemented a sort of "communication" between our plugins. Holger's new shader has direct memory access to emFluid4's internal 3D grids and passes those grids to his particle volume shader. So the only thing that gets rendered are those grids. The particles are completely ignored except for one thing:
The particle volume shader operates only within the bounding box of the point cloud, so you need a few particles in order to have a bounding box that is large enough to hold the 3D grid.
Take a look at the demo scene ".._baVolume..". You will see a little ICE Tree in the Post-Simulation stack that simply adds two particles at the corners of the emFLuid4 fluid box. These two particles ensure that the bounding box of the point cloud is big enough.

Hope that information was useful

Cheers !
Eric

PS: => that means that you must use smaller cell sizes for emFluid4 to get more detail. Note that the demo version of emFluid4 has a fixed cell size of "1".

PS (little trick): using a small cell size in emFluid4 results in longer calculation times... sometimes MUCH longer. Typically, when making fluid simulations, the "interesting" parts come after a few frames, let's say frame 100. With a small cell size you therefore need to simulate 100 frames before getting where you want to.
But there's a trick: you can animate the cell size with emFluid4 ! So simply start your fluid simulation with big/normal cell size, e.g. "1" and then set it to a smaller value a few frames before things start to look interesting.

[Gith]

Thanks Eric!

Yup, i've noticed that particle count doesn't have any impact on the rendering at all. It's great, because it uses a heck less of memory and with trial at home i dont need to worry about 250k particle limitation. I've just use particle for display purposes and then, before rendering i'm just turning them off in Behaviour compound. Right now at work with full version, i'm doing some testing with cell size. I may post some results later if You like.

P.S. Thanks for the tip, Eric!

[Mootzoid]

Hey Gith,

no problem, you're welcome for the tip ;)
And yes, please post some pics !

Cheers,
Eric

[origin]

hmm. I can see that cell size affects quality. So it works similiar to fumefx.
But what about particles?
Are they only for display purposes?
4 particles won't make same image as 4000 right?
Particles passing thru those cells sets some data on them?
What would be the best workflow for this shader, generating one particle per cell for display purposes (like fume)? In the centre of this cell?
Position in this cell doesn't matter?
ba_fluid_value color outputs only velocity for now (color doesnt work)?
so many questions, too bad Holger wrote he wont share any info or support

[CiaranM]

hmm. I can see that cell size affects quality. So it works similiar to fumefx.
But what about particles?
Are they only for display purposes?
4 particles won't make same image as 4000 right?
Particles passing thru those cells sets some data on them?
What would be the best workflow for this shader, generating one particle per cell for display

Particles don't set data on the cells, the cells set data (velocity etc.) on the particles. The grid cells are always there, wheter you choose to visualise them by advecting particles or not.

If you like, you can simply visualise the cells and their data by using the visualize grid cells output from the solver node. Or, you can dig into that node and find the arrays for all the data types. You could use them to set the datas on your own grid pointcloud for preview purposes.

[rconover]

I'm excited to finaly see a promising shader for fluids in Softimage!

I'm curious as to what the render times per frame (ball park) folks are getting?
Another great thing about Fumefx is how fast it renders (for the detail you get)
even compared to regular BA volumes (which I find quite slow with complex/high rez detail).

[Gith]

So.. is there any way to affect cell data? Like add forces, goals, etc. ? Because after doing something with particles it doesn't affect the volume rendering as You've said.

P.S. I'll dig into the sims later in this week, 'cause a lot of things are going on now.

edit: hm, it seems only Emit velocity has any effect on the cell simulation. And it's only SRT based, so it works as an injection of velocity into the simulation from an object source, not point-based. Hm. Any other ideas? If i was to do a tornado for example. Will i be able to do it now, Eric?

[CiaranM]

Well, that would be where the emFluid emit data from SRT or emit data from points nodes would be useful e.g. emit velocity would be something liek adding local forces. What you could also do since we have access the arrays of cell data, is to make another pointcloud from those positions, do something on those points, then feed that pointcloud back into an emit data from points node in the fluid sim. From that you could emit velocity per cell.
But, yeah, it would be awesome if we could write directly to the cells (hint, hint, Eric). Right now that looks to be black boxed.

[Gith]

Oh, ok. I'll do some experimenatation then! Thanks!

[CiaranM]

Also, if you want to to something like a tornado, the best workflow would be to emit fluid from particles. The particles would be simulated in a regular ICE way. You can then emit data from them for emFluid to simulate with e.g. veclocity, density etc.
This is a pretty standard way to do stuff because of the less controlled nature of fluid sims.

[owei]

Hi guys..!

Well, since this hole stuff is really a quick hack, so don´t expect to much for now. I´ve done a quick upload of one of the scenes I´ve modded for my small demo. You can download it here:

http://www.si-community.com/download/sc ... ine_BA.rar

For now it´s black boxed, as Ciaran already said, but using all the emitters an forces inside emFluid4 will take effect on the volume as you can see in this little scene...hope this helps a bit ;)
For modding older emFluid4 setups, dig into the "solver" compound and then into the "solver advanced" compound. There is a port called "fluid buffers pointers". This should be piped into a set data node with "self.__bafluidbuffer" as reference..

cheers,
oliver

[Gith]

Thanks Oliver! I've downloaded Your scene Yesterday and been playing with it some time. But i've noticed that, when You remove the normal emitter it's still the same. It's not affecting the fluid sim. But i'll dig into the compounds as You've suggested.

[owei]

...the velocity emitters are effecting the sim. Both emitters emit velocity..play around with some extreme values (100 or more..) and you will have a obvious effect ..

best,
oliver

[Gith]

Ah, You are correct. My bad. I forgot about that compound attached to the main emitter. Sorry!