FumeFX

#4961

Aaaah that seems to work!!
Fantastic, thank you very much! :applause:

#4962

Cool if it’s works :slight_smile:

#4963

I have this problem as well, it’s so frustrating. I did the following sim at 400x400x400…

https://vimeo.com/43319446

Now, I was very happy with that sim, because I love how the fire looks. Then when I raised the grid res to even 450x450x450, my fire got all wispy/stringy and fake looking. It looked horrible, and I changed nothing but the grid res. Do other fluid simulators have this problem, or is it inherent with Fume?

I’ve been wanting to demo Phoenix, but it seems like almost no one uses it and there are very few resources on how to use it. I really like Fume alot, and would love to get predictable results out of grids over 400.

#4964

Daniel, that is a result of the fluids’ behavior changing because of the higher detail.
In other words it just reacts differently.
I’m currently simming at 1800x630x1360 (I even did 4000x1400x3000) and I can tell you that from a certain scale everything changes. You simply have to rethink the sim.
One parameter which can help is the expansion (Fuel). Up it and the fluids will enlarge and will get back to the volume it was at lower res. Also upping the Temperature buoyancy can offer some help. Of course this al depends on the type of effect you’re after in FFX.

The only way to get it back to your low res behavior is adjust parameters. And it’ll take a lot of test and sim work (at least with me). From a certain scale the behavior is completely different (it varies per effect so you can not put a real border-scale value at it when it will change), although it may be physically correct off FFX(I can not really comment on that).

But on a side note 450x450x450 should be do-able and sim-times should be reasonable.

I can not comment on Phoenix, I only played a little with the demo.

Edit: when you work wit particle sources you may consider upping the radius (in FFX).

#4965

Perhaps one day in the near future, GPUs can accelerate fluid sims fast enough to get manageable sim times at 1000+ grid res. That would be nice to be able to do.

#4966

Yes I would like that to, and at a reasonable price (no Tesla-like prices)

#4967

Well firstly I assume you are using a spacing of 1.0 since you aren’t saying. So based on that you do realize that you are simulating 64,000,000 voxels at 400 cubed and 91,125,000 at 450 cubed.

Dude that is not a moderate increase in voxels, you are adding almost 30 million more voxels to the grid! Sorry but any simulation is going to change drastically when you add that many more voxels to a size optimized grid. If your spacing is even lower then there are even more voxels!

So that is 30 million more voxels for the fluid to travel through and 30 million more calculations for advection, gravity, bouyancy, +100 other params. Just as a matter of understand the scope of what is really happening per step.

#4968

The actual grid dimensions with a 1.0 spacing are 360x360x360.

#4969

Well where do the numbers you mention come from then? (400x400x400) I am confused.

Or are you saying you are dropping the spacing to 0.8 and lower respectively?

LOL not that it matters there are still a shit-ton more voxels in the grid. Even RF behaves differently when you increase the res.

#4970

Yes, exactly. The actual size of the grid is 360 cubed. I lowered the spacing to .9 (400 cubed) and then 0.8 (450 cubed) and the sim changed for the worse.

#4971

The truly only way to preserve the motion you like and up-res is to wavelet. It has its limitations too of course. I do agree with you completely, it is next to impossible to know exactly what is going to happen when you change the density. Experiments and experience are the only way I know of (much like dealing with density in Krakatoa too).

So out of curiosity have you tried running a wavelet on that sim? I can say we beat the shit out of wavelet during spiderman and 85% of the time it held up amazingly well.

Oh and you can break up your initial blob with some simple geo that interferes with the path of the fluid. I like using a torus knot with fairly low steps and 3-6 sides. Of course having this deflector geo in the way will increase you sim times but should add nice variation. You can also use this as an object source and gradient map velocities to it as well :wink:

#4972

I am working on a project where i have to replicate the dust falling off the character; ground impact dust…and the smoke that emnates behind him…i have tried a couple of settings and techniques in fumefx but i am unable to get the desired motion of the smoke…any helpl would be appreciated
also i was not getting the soft an wispy dust look…(was using pf flow sourced fumefx)
shot-http://www.youtube.com/watch?v=qVLPD30yF_4&feature=youtu.be

#4973

John,
Just to be sure: Wavelet only adds more details “in” the effect, but doesn’t really affect artifacts due to too low resolution (not enough voxels, too high scale), right?
The artifacts are mostly noticeable at the outer edges of fire (jagged edges). Just wondering, since I’m really pushing my hardware and FumeFX to get fine detail at the edges (and to a lesser degree also in the fire). The only way for me to do this, is to increase the amount of voxels.
Or are there any tricks with wavelet you can pull of to influence this? Any experience you have in this area I would love to hear about! Also what the 15% of cases are where Wavelet didn’t help you :wink:

Best regards,
David

#4974

In my experience with wavelet, it tends to add too much turbulence. I usually dial back the strength to 1.0, and threshold to 0.5.

#4975

Yes it does get rid of voxel step artifacting. (EDIT: Let me be more correct, it doesn’t get rid if it it just makes it smaller.) Most if not all of it, you are basically subdividing the saved voxel and running a wavelet algorithm on each of the subdivided data (not sure what the true programmatical terminology is for it).

Certainly depending on the look you want you will want to adjust the Strength/Threshold values. As a rule of thumb I like stronger effect for smaller scales and less effect for larger scales.

Just a quick visual example (download the original)

https://vimeo.com/43828606

#4976

Thanks for the quick replies, guys!
In the beginning i did some quick wavelet tests, and disregarded as a high res solution, because it didn’t seem to solve the voxel artifacts. I was too quick in my judgement it seems.
I’m gonna test wt straight away. I am very curious to see the result on the fire projects.

#4977

In my experience, wavelet turb adds very nicely to flames indeed. Here is one I did about a year ago, and I ran it at fairly low res around 120 voxels I believe, and added 2.0 wave turb…

http://www.youtube.com/watch?v=ssK9RiIkpM8&list=UUf2UVMAUbiUNTuRQo7Gl6RA&index=3&feature=plcp

#4978

That is very nice fire!

#4979

QFA :slight_smile:

Where is that surface smoke coming from, or is it fuel set to renderable? It looks kinda cool

#4980

Well, that fire was made directly from a tutorial by Allan Mckay in his Fume FX Core Fundamentals DVD. It involved having particles shoot fuel inward toward the emitter center instead of outward. Those fluids colliding would then give you those flame like wisps.

However, I no longer do flames like that. The reason is the particles give away their ignition point. I now emit from a noise map on the surface of the object I want to burn. An emission noise in fuel and temperature, with the right sim settings, can give you very photoreal flames.

I have uploaded an example file if you guys want to see how I do my flames. Keep in mind this is set up with linear workflow. The lower the spacing, generally the better the flames will look. It sims pretty fast anyway. Here is the link to the file on Dropbox…

https://dl.dropbox.com/u/73226790/flame_line_02.max