Some fluid animations for a paper

I just finished an animation made with elbeem and Blender for a paper, and thought it might count as a “finished project” :slight_smile: The setups themselves are not that interesting (the obligatory falling drop, breaking dam and glass of water) but the simulation are largest ones I’ve done so far, and I think the breaking dam (first half) and the glass one turned out well.

Here’s the whole animation as Divx:

And as lower quality Mpeg:

Actually, the later frames of the breaking dam animation took over 24h to render with Yafray on a 2.2GHz Opteron. I had to set the ray depth to 11 for photon maps and rendering, as lower values produced overly dark regions. Without the two side glass walls it probably would have been a lot faster, but in this case I think it looks better with them. The whole breaking dam simulation there took a week on a quad-opteron… The last frames of the 3rd glass animation also took ca. 14h btw. In this case the simulation was “quite fast”,it still took a few hours.

as from my point of view (blender newbie, but student of cg :D) this looks very good.

Is it possible to get a copy of the paper, or is it still some sort of “top secret” ?

Whoa, that was some impressive demonstration.

The MPEG is just a title/description page, perhaps a sixth of a second long.

Really nice, i think that the one that is “filling” the glass cup is too violent, but other than that they look really nice

thanks dude, ppl like you makes blender greater everyday

i have a problem with that video…it looks real…but can i drink it? it looks pretty tasty…but you should have rendered it with some flavor…

Very good, well done fluid… it must have took ages to bake…

The MPEG is just a title/description page, perhaps a sixth of a second long.

I had that problem too. Download it with Save As, it worked for me.

Oh, so YOU’RE the guy who added this fluid simulation stuff to blender!

:smiley: :smiley: :smiley:


grassi3000: we’ll publish the paper as a preliminary tech report or so, I’ll post the link once it’s up… here’s just the abstract:

Blender is explicitly mentioned to demonstrate the whole thing :slight_smile:

Correct me if I am wrong…

Were those 480^3 domain simulations done in Blender?? How much RAM do you have? I have never seen Blender used for such a high resolution sim before…

EDIT 1: They were done in Blender/Yafray! (It says so on the mediawiki) Wow! How long did it take to simulate then?

EDIT 2: It also says on the mediawiki that “Currently the resolution is limited to 200 (ca. 850MB)”. I’m confused. Sorry for all the edits.

Anyway, a little message to n_t:

Your work and the algorithm you described were enough of an inspiration for me to code a basic LBM simulation as a C project for my physics degree. I got it marked today for 4 days practical (1/3 of the practical course this year!).

So thanks for your help and for the inspiration.

Oh and thanks for coding this into Blender too! :smiley:


P.S> Keep up the good work! I’m looking forward to moving collision detection with meshes.

P.P.S> Was it based on my suggestion that you added inflow and outflow? :smiley:

koba: Yes you’re right, the simulations here have really high resolutions. They were calculated with same simulator, though on a Quad-Opteron node (64bit linux) with 16GB of memory. But those two test cases were merely there to show off the high resolutions, it wasnt really necessary in this case. I also updated the wiki, the resolution is now limited only by the amount of memory you have (or your operation system allows you to have). And cool that you coded an LBM simulator. But I don’t exactly remember what/who made me add the in/out flow stuff :slight_smile:

I got a request from Vassilios about the .blend file for the glass simulation, perhaps this is also interesting for s.o. else:
Note that the settings are all for high quality, so simulation and rendering with Yafray will take quite some time… Btw. I the animation a while ago, and didnt have time to really test the file again, I hope it’s the right version…

PS: grassi3000 - the paper’s still not online, but hopefully will be soon,




At last the pre-print version of the paper is up… I think it’s the most complete description of the algorithm that’s used in elbeem so far (and hopefully more or less understandable :))

thx, that looks very interesting :smiley: