PDA

View Full Version : PhysX fluid simulation!



danielkaiser
04-23-2013, 08:52 PM
http://physxinfo.com/news/11109/introduction-to-position-based-fluids/

Greenlaw
04-24-2013, 02:50 AM
Want. :)

Kaptive
04-24-2013, 09:56 AM
http://physxinfo.com/news/11109/introduction-to-position-based-fluids/

Hah! You beat me to it. I just came here to post it. I'll post the video in below though so people can have a look at it quickly before reading further.


http://www.youtube.com/watch?feature=player_embedded&v=6WZZARzpckw

So, I know little about programming... but........ Isn't it possible to bring the PhysX calculations and effects into 3D software such as Lightwave? Bring it in and bake it, from the perspective of incorporating this into the Lightwaves program architecture. Is there some limitation brought about by incompatable data or something? It seems such a pity if so.

A PDF about it from PhysX below...

Position Based Fluids PDF (http://mmacklin.com/pbf_sig_preprint.pdf)

Kaptive
04-24-2013, 09:58 AM
Oh, and one other video of it...


http://www.youtube.com/watch?feature=player_embedded&v=F5KuP6qEuew

Hail
04-24-2013, 10:25 AM
WOw!

Ernest
04-24-2013, 08:36 PM
OK, anyone who marveled at DNT back in the 90s answer honestly: back then did you really expect to see this level of fluid sim -in real time- in your lifetime?

JamesCurtis
04-24-2013, 10:05 PM
Gotta admit, nice stuff, and sims like this in less years than I'd imagined!! Only thing about the sims to date is that most materials will tend to retain a little fluid, maybe even just droplets clinging to them. Soaking action? This really needs to be figured out too!!

Kaptive
04-24-2013, 10:45 PM
Gotta admit, nice stuff, and sims like this in less years than I'd imagined!! Only thing about the sims to date is that most materials will tend to retain a little fluid, maybe even just droplets clinging to them. Soaking action? This really needs to be figured out too!!

lol... never satisfied :P

Ztreem
04-25-2013, 12:36 AM
Looks very useful, put it in LW so we can play! :)

Netvudu
04-25-2013, 04:48 AM
Personally I think itīs a pretty poor simulation. I understand itīs real time, but the amount of detail there is very low. That would never pass for water in any film. Maybe in a game it would be acceptable...

Ztreem
04-25-2013, 05:54 AM
Personally I think itīs a pretty poor simulation. I understand itīs real time, but the amount of detail there is very low. That would never pass for water in any film. Maybe in a game it would be acceptable...

I never seen a game with that kind of water simulation. For a Hollywood blockbuster maybe it's not detailed enough but for an animated short or cartoon it's ok, I think. They also state how many particles they use in each simulation so you could asume that you can increase the number of particles but then ofcourse it will affect the fps and stop being realtime. If you compare this to anything we have in LW today I think its really good.

vncnt
04-25-2013, 05:55 AM
If LW could only half of that.
I'm in love.

Greenlaw
04-25-2013, 06:33 AM
Personally I think itīs a pretty poor simulation. I understand itīs real time, but the amount of detail there is very low. That would never pass for water in any film. Maybe in a game it would be acceptable...

Wow, you're tough. Personally, I can think of dozens of films/commercials where I could have used this in the past and the director and producer would have been absolutely thrilled to get it. A program like RealFlow can make them nervous because it can take a day or two to get a sim calculated, and sometimes a few days is all you get. Besides, just because these demos are presented in realtime doesn't meant you can't allow it spend more time to achieve higher quality sims--I would rather spend minutes or even a few hours getting a good sim than waiting a day or two for results.

G.

Kaptive
04-25-2013, 01:24 PM
Any programmers out there know if PhysX data can be accessed and potentially utilised within Lightwave? Anyone from Newtek care to comment?

As said above, it might not be realflow, but I think it would be more than capable for application in a large number of productions. Fluids are one area that could give Lightwave a real edge. If it were possible to utilise PhysX for this (with an Nvidia card of course)... there would be no need to reinvent the wheel.

cagey5
04-25-2013, 01:26 PM
Isn't Steve Hurley using PhysX in his beta dynamic engine for Lightwave?

creacon
04-26-2013, 06:28 AM
Look for a few of my threads in the plugin/developper section ;-)

http://forums.newtek.com/showthread.php?133662-Particle-Buffers

I can't find the other thread but I remember I called it Splash.

And now give me 48 hours in a day and I'll finish it.

creacon

Greenlaw
04-26-2013, 08:45 AM
** Deleted **

Whoops! Sorry, posted that to wrong thread.

geo_n
04-26-2013, 09:10 AM
Wow that's very nice. It would be great having this built into lw. I did a lot of commercials for beer and softdrink companies that could use this kind of quality.
Those companies are corporations that don't go broke unlike film companies. Rather target that market for newtek.
I used blender for those commercials but if I could have that in lw...

Netvudu
04-26-2013, 06:40 PM
Wow, you're tough. Personally, I can think of dozens of films/commercials where I could have used this in the past and the director and producer would have been absolutely thrilled to get it. A program like RealFlow can make them nervous because it can take a day or two to get a sim calculated, and sometimes a few days is all you get. Besides, just because these demos are presented in realtime doesn't meant you can't allow it spend more time to achieve higher quality sims--I would rather spend minutes or even a few hours getting a good sim than waiting a day or two for results.

G.

I totally get your point, and it makes a lot of sense. The problem being, after spending A LOT of hours with fluids simulations, many of them in Realflow (nowadays itīs mostly Houdini although Iīm still 100% up-to-date in Realflow latest versions as well) I can tell you one thing for sure: It doesnīt matter how cool your motion gets or how many times your refine your sim, once you reach a certain "acceptable" solution, the only thing thatīs going to improve the detail and realism of the sim will be more particles. At around 2 million it starts getting interesting, and nowadays you can sim about 10 million quite comfortably in a few hours on any half-decent system or even laptop.
Meaning that this kind of real-time solvers would move you faster into the "acceptable" range, but then they wouldnīt be able to reach that far as more mature solvers, and the real time ones tend to scale terribly once you start to crank up huge numbers.

Iīm not trying to be tough. I really do think those simulations look terribly fake. If it wasnīt for the real-time factor I would consider them mediocre at best.
Then again, Iīm sure one day we will be handling huge amounts of data in real time....who knows what new features we will have then to allow us to bend our systems to theri knees again and not work in real-time heheh.

Titus
04-26-2013, 06:57 PM
Any programmers out there know if PhysX data can be accessed and potentially utilised within Lightwave? Anyone from Newtek care to comment?

As said above, it might not be realflow, but I think it would be more than capable for application in a large number of productions. Fluids are one area that could give Lightwave a real edge. If it were possible to utilise PhysX for this (with an Nvidia card of course)... there would be no need to reinvent the wheel.

D-Storm has (had?) a fluid sim plug-in using PhysX. They should know how to use this, but are they interested?

prometheus
04-27-2013, 06:54 PM
D-Storm has (had?) a fluid sim plug-in using PhysX. They should know how to use this, but are they interested?

To bad Dstorm donīt have a proper english page for it, it calculates quite fast though, Not sure that guy from dstorm is lurking here, might be completly unaware of it.
http://www.dstorm.co.jp/dsproducts/plugins/LiquidPack/index.html
http://www.dstorm.co.jp/dsproducts/plugins/LiquidPack/naminamifx.html
http://www.dstorm.co.jp/dsproducts/plugins/LiquidPack/particle.html

Areyos Alektor
04-27-2013, 07:01 PM
The Physx fluids is nothing new, it's the old system from Ageia. I still have my PPU running on an old PC to run small homebrew programs. But no NVIDIA has remove the management of the PPU in the latest drivers :(

Kaptive
04-27-2013, 08:10 PM
I totally get your point, and it makes a lot of sense. The problem being, after spending A LOT of hours with fluids simulations, many of them in Realflow (nowadays itīs mostly Houdini although Iīm still 100% up-to-date in Realflow latest versions as well) I can tell you one thing for sure: It doesnīt matter how cool your motion gets or how many times your refine your sim, once you reach a certain "acceptable" solution, the only thing thatīs going to improve the detail and realism of the sim will be more particles. At around 2 million it starts getting interesting, and nowadays you can sim about 10 million quite comfortably in a few hours on any half-decent system or even laptop.
Meaning that this kind of real-time solvers would move you faster into the "acceptable" range, but then they wouldnīt be able to reach that far as more mature solvers, and the real time ones tend to scale terribly once you start to crank up huge numbers.

Iīm not trying to be tough. I really do think those simulations look terribly fake. If it wasnīt for the real-time factor I would consider them mediocre at best.
Then again, Iīm sure one day we will be handling huge amounts of data in real time....who knows what new features we will have then to allow us to bend our systems to theri knees again and not work in real-time heheh.

But surely, the point is that currently in Lightwave, there is no solution at all to get something even 20% of the way to what we see here. I understand your points with regard to quality, and getting that ultimate look, but not everyone needs absolute photo realism. If they did, then that is when you would turn to 3rd party such as realwave.

However, a simple fast solution to achieving acceptable results in Lightwave would be most welcome in my humble opinion, and that would be all I'd expect of it.

Kaptive
04-27-2013, 08:17 PM
D-Storm has (had?) a fluid sim plug-in using PhysX. They should know how to use this, but are they interested?

Well so long as you can get data out of PhysX, then perhaps someone else will have a go at it at some point in the future.

I noticed on the Dstorm site that they have something called Liquid Pack. The results (from the small pictures I could see) didn't look too convincing, so a revamp of that with an updated fluid system might just happen. It seems to utilise Nvidia cards from what I can make out. Here is the google translate link...

http://translate.google.co.uk/translate?sl=ja&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&eotf=1&u=http%3A%2F%2Fwww.dstorm.co.jp%2Fdsproducts%2Fplu gins%2Findex.html&act=url