Smoke and fire

[starbase1]

What is the current view round here on 3rd party smoke and fire type plugins? I'm toying with the idea of getting something but pretty much everything I look at seems stalled, with questions over if they are still active products...

What are people using that they are happy with?

I seem to recall that 'dynamite' was a bit wobbly, but note that Napalm is now in my price range...

Nick

[Phil]

There is no straightforward answer :

Napalm is just a particle system, and one which the developer seems to be struggling to maintain. They still don't have a timeline for a UB port of any of their plugins and there seems to be a single developer who is attempting to figure out what needs doing. *shrug* It may be cheap, but if the developer isn't maintaining it, you may well find that once 9.x breaks it, you won't get it running again.

Dynamite meanwhile brings fire/smoke dynamics, and a volumetric engine. The problem there is that the developer hasn't shipped a formal update since the product was first released. The 1.2 update has not been released and you have to work very, very hard to get a response from the developer. The 1.1 version that is available is, well, a little buggy (multithreaded rendering is hugely problematic). PFX-driven volumes are also often slow to render.

Dynamite is full of potential. If the developer was responsive, and actually shipped updates rather than trying to perfect everything in a single iteration, I'd be more comfortable recommending it. The likelihood is that you will find it very useful right up until you hit a problem. Then you will go blue in the face trying to get the developer to respond and issue a fix.

[starbase1]

Thanks for the info, pretty much what I had figured...

A souped up particle engine may be very useful to me in it's own right - I'm always hitting the particle limit with the built in system.

Any thoughts on Pawel Olas' fire and smoke?
Nick

[Phil]

Nope. No experience.

[jasond]

I've used it. It's a great little plugin, but it's just a particle emitter, so you'll encounter whatever limitations HyperVoxels bring to the issue. But the built in particle motion and controls in F&S are excellent. I think you may get serviceable results from HVs with a lot of patience.

[starbase1]

I've used it. It's a great little plugin, but it's just a particle emitter, so you'll encounter whatever limitations HyperVoxels bring to the issue. But the built in particle motion and controls in F&S are excellent. I think you may get serviceable results from HVs with a lot of patience.

Does it share the same particle limit as bog standard Lightwave?

[Phil]

One other option might be pyro (same folks as Napalm), although I suppose it really depends on what you are looking for... Pyro is, I think, a pixel filter. That means it won't reflect/refract and it will have the usual issues with volumetrics.

[starbase1]

Perhaps it would help if I explained some of the ingings I want to do.

I have never got rocket exhaust I liked with the built in stuff. Not even close. Attempts to make it better seem to drive me rapidly to many more particles than native LW tools will handle.

I've also had no joy in trying to make the edges slightly soft of clouds or smoke, I either end up with stuff that looks more like cotton wool, or the whole lot is transparent, getting control of edge effects with HV's completely defeats me. And that's just for a simple 1 jet type image. For a ring of 30 engines under an N1 moon rocket, each with a long thin plume 150 metres long, I don't know where to start to even make a bad one...

I do also like the idea of the realistic motions, the clips of "fire and smoke" with things moving throuh the smoke and disturbing it look a lot of fun.

At it's most basic, increasing the lparticle limit 100 fold would let me get further I think... (What drives that limit? It's been there for ever, and with the amount of memory on a modern machine I wouldd have thought it could have increased over the years...)

Nick

[jasond]

Does it share the same particle limit as bog standard Lightwave?

I used it to mimic some fake airflow through a room and it worked fairly well with 5 thousand particles, if I recall correctly.

[Mr Rid]

I've also had no joy in trying to make the edges slightly soft of clouds or smoke, I either end up with stuff that looks more like cotton wool, or the whole lot is transparent, getting control of edge effects with HV's completely defeats me.

Try a local density gradient in the dissolve to soften voxel edges. I use gradients in most channels for density, age or speed. A rocket exhaust should be easier than other types of FX Ive been able to get decent results for. But a mistake most CG artists make is trying to pull it off entirely in 3D. A post tool like AFX or Fusion can help greatly to more quickly take a rendered element further than you ever can in 3D.


I do also like the idea of the realistic motions, the clips of "fire and smoke" with things moving through the smoke and disturbing it look a lot of fun.

I dont see anything in the Fire & Smoke demos that can not be done in native PFX, although it may take more steps to set up swirly smoke behavior (tumble some vortex or donut winds with no falloff).


At it's most basic, increasing the lparticle limit 100 fold would let me get further I think... (What drives that limit? It's been there for ever, and with the amount of memory on a modern machine I wouldd have thought it could have increased over the years...)

Nick

What limit are you talking about? You can use a million particles but you shouldnt need anywhere that many. Study actual reference closely.

Its odd but LW's 'dumb' particles (not self-colliding) calculate faster than RealFlow's. But you can forget about high counts (choked at about 25k for me) of self colliding particles in LW.

[Andyjaggy]

I've had a lot of success using the vortex wind to create smoke. It seems to give a really great natural movement that seems inherent to smoke.

Also try rendering out a volumetric and a sprite pass and comping them.

[starbase1]

What limit are you talking about? You can use a million particles but you shouldnt need anywhere that many. Study actual reference closely.


I'm going from memory here as I am at the day job, but I thought the limit for a particle emitter was about 10,000?

And can you be a bit more specific about a gradient in disolve please?

That sounds promising.
Nick

[prometheus]

I just have to agree with Mr Rid, I think you can acomplish mostly anything with native particles and hypervoxels just as the pavel olas smoke and fire plugin.
the smoke and fire plugin has some more build in swirl and turbulence settings, but if you learn how to use wind objects properly, thats the way to go. the smoke and fire has emissions from weightmaps wich native particles don´t but using texture maps are almost equal perhaps.

I think you should forget about pyro..seemed reallly slow both in calculations and rendering and no longer support for it.

well currently theres no real shortcut to get that nice sweet hypervoxels with a blink of the eye, it takes time to learn to get good looking and a lot of time to spend to experiment with gradients and all the channels, just listen to Mr rid and test the dissolve channel with different gradients, local density is good.

Since I recently got an reply from sitni sati about fumefx for Lightwave, it seems that we can forget about that for a long time theres no plans for that at the moment.

The most realistic smoke plumes perhaps is if you should choose to go with Dynamite, however it has taken a long long time since the last release and there will probably be some quality issues and issues when using it with big scene elements that covers a wide range when using fluids.

otherwise the dynamite fluids are the best looking stuff I think, and its pretty easy to tweak the fire and smoke after the calculation, it is still missing some more advanced material injections thou. and getting the simulation speed to slow down in a particle age manner.

You can of course use hv sprites but it´s not as good for thicker smoke, you can by using gradients get a decent look perhaps but it will miss in lighting quality and the deeper thickness and shadows.

I have attached a mov sample of hv sprites smoke just a test thou..smoke from a jet shouldn´t look like this thou..just testing.

[JamesCurtis]

I know this isn't a 3D or LW plugin, but I've had occasion to use a 2D program callled Particle Illusion 3 at times for effects over my LW work. It has tons of presets and is quite a complex program for what it is.

Here is the link for it:

http://www.wondertouch.com/

The program is about $300 list.

Just thought I'd mention it.

[Mr Rid]

I'm going from memory here as I am at the day job, but I thought the limit for a particle emitter was about 10,000?

And can you be a bit more specific about a gradient in disolve please?

That sounds promising.
Nick


There was a 100,000 limit up until 9 I think. No longer.

Add a local density gradient in the dissolve. Make the end value 0. Also as Andyjaggy mentioned, rendering a separate pass of sprites to overlay in comp will greatly reduce the puffball syndrome.

Start with Very Low render quality and shadow quality. Turn off Texture Shadows (Ive never seen an advantage to this render hog). Use Viper and flip between Particle and Object preview to see how value changes are affecting. Start simple with lower particle counts and play with different gradients. Lower Thickness (with high smoothness) is the main softening tool but increases render times. First get a HyperTexture you like (Turbulence is a starting point) or start with one of the presets and look at how each value is affecting the look and start playing with them. If you have not checked before, there are also several basic HV and PFX scenes that come with LW that may serve as good starting points.

Particle Illusion can sometimes do very nice simple X,Y moving smoke trails/plumes but had always found it very restricted when tracking it to anything moving much in 3D, particularly on the Z.

[Mr Rid]

P.S.

I usually render smoke FX at no more than 640x, and rarely more than half whatever the final res of the rest of your elements. 2K renders only make voxels look hairy and require softening in post. Why waste major render time only to wind up having to blur it down? Have even used smoke rendered at 1/4 the final res that looked fine. If you render smoke at 640x and scale it up in comp to fit a 2k plate, the natural dithering of the transform will soften the smoke nicely and cut render times by 1/8.

[starbase1]

I know this isn't a 3D or LW plugin, but I've had occasion to use a 2D program callled Particle Illusion 3 at times for effects over my LW work. It has tons of presets and is quite a complex program for what it is.

Here is the link for it:

http://www.wondertouch.com/

The program is about $300 list.

Just thought I'd mention it.

Thanks, I took a look some time ago. I thought it was pretty, bt really didn't like the way it worked.

[starbase1]

P.S.

I usually render smoke FX at no more than 640x, and rarely more than half whatever the final res of the rest of your elements. 2K renders only make voxels look hairy and require softening in post. Why waste major render time only to wind up having to blur it down? Have even used smoke rendered at 1/4 the final res that looked fine. If you render smoke at 640x and scale it up in comp to fit a 2k plate, the natural dithering of the transform will soften the smoke nicely and cut render times by 1/8.

Ah, now that really could be a problem. While I will animate at some point, my main target for the rocket project is very large prints.

[starbase1]

There was a 100,000 limit up until 9 I think. No longer.


I have an FX Emitter panel in front of me, version 9.2, and the particle limit is 10,000. The slider will go no higher.

What am I missing?

[starbase1]

I have an FX Emitter panel in front of me, version 9.2, and the particle limit is 10,000. The slider will go no higher.

What am I missing?

OK, boy do I feel dumb, I can enter numbers manually much bigger than the slider will go!

Well, I can certainly take things a lot further than I could with 10,000 particles shared over 30 rocket nozzles! That left me with about 350 particles per stream, which was a mite confining to say the least!

Thanks Mr Rid, time to stat experimenting!

[prometheus]

He..I guess we have all felt dumb at certain stages trying to learn apps, don´t you worry about that.
And besides that, why shouldt´n we be able to raise the slider to what ever particle count we want? maybe thats just an old thing.

By the way heres a link do download plugs for lightwave when using
particle illusion the plugins are free and they record the motion or position of selected object or null or even get a single particle data by using edit fx node and read path and add a null to that particle.

http://www.wondertouch.com/downloads_other.asp

so just render out your animation to an image sequence and import your motion data from lightwave and attach particle illusion emitters.
Would be nice if there was a major update to particle illusion with true 3d and
some procedurals added to get a more natural noise look as a complement to just image sprites.

Ohh..so right now your main focus isn´t animation but high res prints..oki doki.

[Mr Rid]

Ah, now that really could be a problem. While I will animate at some point, my main target for the rocket project is very large prints.

Scaling up from half or lower res should not matter when using any decent transform tool, especially if you are only rendering a print still that can be doctored in PShop. Smokey elements are very forgiving. Ive even shot many a smoke element on SD video, then deinterlaced, scaled and comped into 2K film where no one could tell.

One place insisted I render a misty white water HV element at 2K, 33 pass which wound up taking 3 days a frame to render! I knew they were nuts to waste time on this. The next similar misty water element they had me develope, I secretly rendered at 512x, scaled and pre-comped as I usually do. It went to the compositor and onto film which was ultimately blown up to 4K for a ride film. The only comments I received from the supers and producers was how good that element looked.

...........................

Years ago, I did MANY CG test renders with various resolutions, aliasing, moblur and the like, and viewed them on film projected onto a large screen to see what all you can get away with. A photoreal element rendered at 640x and blown up to 2K can actually look surprisingly fine. A little soft but no artifacting. The lower rez became apparent mainly when the shot was cut next to any shot originating at a higher rez. But at one place we rendered all full CG shots at 1K (scaled to 2K) which diffused the shots nicely, and only rendered at 2K where the element had to integrate into a 2K plate.

I learned that 9-pass AA worked just fine for most final CG that doesnt have too much moblur. Most places tend to insist on rendering everything with 33 passes when it doesnt really need it. They just never learned where the exact difference becomes apparent. I also always render with adaptive sampling at '.03' to decrease render time without any noticeable difference in AA (even upon close inspection of alpha edges).

I also learned there is absolutely nothing wrong with jpegging a 24 bit texture, element or source plate once at the highest quality setting. Ive had a post super argue heatedly with me about this but I know that jpeg artifacting at the lowest compression setting is extreeemely minor and only noticeable when zooming into pixels in dark areas where there is a very very tiny difference that will never be apparent in a final comp.

[DLIT]

Is it not possible to achieve realistic results with LW HVs ? Would NT be kind enough to set up a tutorial about this, possibly with the help of William Vaughn who released so nice video tutorials recently bringing big help to people like me and whom I take here the opportunity to thank for it.
Dom

[prometheus]

Is it not possible to achieve realistic results with LW HVs ? Would NT be kind enough to set up a tutorial about this, possibly with the help of William Vaughn who released so nice video tutorials recently bringing big help to people like me and whom I take here the opportunity to thank for it.
Dom

of course it is possible to get very realistic and convincing smoke and fire..
Mr Rid has some of the most convincing stuff what I have seen on the forums, check out his samples fireball and flamethrower so try and follow he´s suggestions, He´s results are probably due to lots of tweaking and learning under years, not something that everyone pulls of within two or three weeks.

But of course It will probably take time to learn and how fast and easy it is to set up depends on how much time you spend on it and checking real reference I guess.

You will probably get going faster with fluids like dynamite, but even that has limitations.

[Mr Rid]

HV and PFX have been developed very little in years and seem to low priority for NT. Unfortunately the current tools do not make realistic FX easy to achieve. I have just inadvertently found myself one of the few masochistic enough to spend too many hours strangling HV into something useable, but that always requires help from post processing.

But HVs dont necessarily need a major renovation. Ive posted many times that there is one relatively simple HV feature addition that would go a long way in achieving more convincing FX- proper Distance Between Particle gradients in the blending and particle size. This would eliminate the currently prevailing 'puffball' look.

Have communicated with NT about providing tutorials, demo shots and examples scenes for the next LW version but the discussion didnt get far. So perhaps Vaughan will offer up something.

With a seemingly missing developer, I cant hold my breath for Dynamite at this point. FumeFX has no plan for LW. RealFlow is only good for liquids. So there appears to be little future for FX work in LW(?).

[prometheus]

Yepp I agree with you to about it takes a lot of time to get hv good looking stuff and some postprocessing.
I also feel It´s taken to long time since a overhaul of the hypervoxels was done, I do hope thou that this soon will be adressed more, Even before the Lightwave 9 cycle is over.

Even thou I really love the fx stuff with particles and hv:s, I also think Newtek Has focused on the right thing thou, It´s only the speed of the development that makes me somekind of frustrated to watch but I can Imagine the kind of hard work and challenges the newtek team was facing when rewriting so much of the program, and still are.

with the new lw 9.5 coming up they seem to have followed up on the character improvements and throwed in a hair solution and those things has been wanted from a lot of people as first prioritys.

along with some further enhancements it looks good, However I thought they
should have reached further in the porting of the model process With somekind of enhancements on getting many model tools parametric with history and perhaps some kind of node procedural approach a la houdini.

so after they get the model process done, the full attention might be on
particle fx and dynamics and volumetrics Or maybe they can pull all those things of at the same time.

I am also a bit worried about the development of Dynamite since it was so long ago it was released and nothing seems to happen in that area, so we are all left with speculations.

The tech stuff, distance to particle sounds good, I really don´t know how this works In other programs, If it is the size or stretching to nearest particle or
if there is a Interpolation particle filler of somekind. or maybe there also should be a subframe calculation for lw particles.

yeah fume fx are not an option for lightwave Ivé been told.

I would love to find out how Houdinis volumetric and image 3d system looks and renders,I have the apprentice version but I really hate setting simple sprites up in Houdini and I just can´t get any grip on working with voxels or volumetrics in houdini, so in that regards I love the workflow and speed to set up hypervoxels in lightwave.
I will try Houdini again someday but it really is very different to learn compared to most other programs.

[starbase1]

I've managed quite reasonable fire before now, (and it's for download on mmy pages), convincing fast moving churning exhaust was more of a problem.

But I've started experimenting with the much higher particle counts, and it's already looking a LOT better. Once I get some rough edges removed all start a WIP thread with all the stuff for download, so others can join in.

Part of the stall is that I am re-thinking what an N-1 exhaust actuallty looked like as the video I have is not great. I recently discovered it's the same fuel as a Saturn V first stage despite appearing very different in the film, (wildly differiung exposiures also don't help!).

[DLIT]

Here is a link to a site dedicated to LW tutorials that some may know, others not. Useful tuts about HV can be found there. Hope this helps.
Dom
http://members.shaw.ca/lightwavetutorials/Newest_Links.htm

[Mr Rid]

...

The tech stuff, distance to particle sounds good, I really don´t know how this works In other programs, If it is the size or stretching to nearest particle or...

Right- http://www.box.net/shared/static/ietmqz76vh.mov
The Blend in HV just needs texture tab with a global Distance Between Particle gradient option available. Currently there are separate X,Y,Z 'DBP' gradients that have an impractically sharp falloff and dont work at all in most channels. There is also no DBP option in the Size texture where it would be the most useful.

Current problem-
http://www.box.net/shared/static/m1sqicrwo4.zip
This simple scene has single particle emitters aimed on the + and - X, demonstrating how each channel's X axis DBP behaves. May select any instance in HV and run a quick Viper Scene preview to see. What is odd is that the top 4 examples render with no effect, but they do in the GL(?).


I would love to find out how Houdinis volumetric and image 3d system looks and renders,I have the apprentice version but I really hate setting simple sprites up in Houdini and I just can´t get any grip on working with voxels or volumetrics in houdini, so in that regards I love the workflow and speed to set up hypervoxels in lightwave.
I will try Houdini again someday but it really is very different to learn compared to most other programs.

I know a few Houdini artists (one of them did the famed 'BAMF' effect in Xmen) and Houdini is definitely the particle system, especially now that it has fluid dynamics. But the Master edition is priced only for businesses, and as one LW/Houdini artist explained to me it was very complicated to get it to interact with LW scenes since it isnt directly supported- jumping thru various conversion hoops. And as a staunch right-brainer, I dont want anything to with expressions, scripting or code of any kind. I wanna drive the race car fast around the track and leave the inner engine workings to the left-brained mechanic. Would rather be 3PO and leave the tech R2.

I can understand NT putting off aspects of development where there is a decent plugin or alt software to fill in the gap. But if Dynamite has stagnated, then I think you have to LW behind to do serious FX work.

[Mr Rid]

Here is a link to a site dedicated to LW tutorials that some may know, others not. Useful tuts about HV can be found there. Hope this helps.
Dom
http://members.shaw.ca/lightwavetutorials/Newest_Links.htm

Yes! Some are old but it looks like there are a thousand LW tutorials there
http://members.shaw.ca/lightwavetutorials%20/Main_Menu.htm

[prometheus]

yeah Mr Rid, that hv blend sample shows how simply sweet it could be with distance to particle options especially sweet for fluids.
Lets keep on with the nag about this, someday they might wake up:)

Heres.
a quick sample pic over the open gl of some particles in houdini vs lightwave.
setting that up was a bit easier in houdini and the driving the particle sim in houdini with turbulence wind amplitude noise is easier than using a wind and
procedural textures in the velocity channel in lightwave.
so a better open gl display for controling particle size an color and better or new type of noise fields would be great apart from some improved particle amount handling of course.

D-storms jet stream wasn´t perfect but promising but that one hasn´t been updated for a long time either, it had some nice particle size controls and doublers..there should have been som kind of particle doubler/interpolator where particles breaks away from eachother at a given distance, for example distance to particle, that way it could render out much more particles than calculated and with control over where it breaks instead over the whole area.

The bamf effect was really sweet and just to become famous.
a lot of blends and tools for that one I guess both maya, and partman and houdini and mantra renders. something about 20 millions of particles. heres the topic from computer graphics world...Quote
"To create the effect, the Cinesite team started with a rough model of actor Alan Cummings that was fashioned using Alias|.Wavefront's Maya in six sections: head, torso, arms, and legs. Then, in Side Effects Software's Houdini, the crew filled the virtual volumes with particles that were rendered in Side Effects' Mantra using a smoke shader and Houdini's i3D volume rendering datasets. With the body in sections, the artists could control each volume to make him dematerialize all at once or have an arm or a leg lag behind. To make the smoke disappear, they pulled particles in the volumes inward and rendered the implosion with Maya.

The streaming smoke and the residual smoke were also created with particles and were animated with a fluid dynamic algorithm written by Jerry Tessendorf that was ported into Houdini as a particle operator and used as a turbulence field. To render a nearly impossible number of these particles�sometimes as many as 20 million�they used a proprietary tool called PartMan, written by Tessendorf and Bill LaBarge, 3D technical director. "We could render six million particles using 125mb of RAM and a general pass in 15 to 30 minutes per frame at high resolution," says Anderson. "This meant we could render the effects in multiple layers and recombine them in the comp." The compositing was done with Kodak's Cineon software.

To fasten the smoke to the live-action Nightcrawler, the compositing crew matched-moved the rendered output into a relative position, rotoscoped Nightcrawler, and then replaced his alpha with the render's alpha. "On top of that we did a series of 2D distortion effects," Anderson says. "As he implodes, we gave it a slight expansion outward before the effect and then collapsed it inward."

[Mr Rid]

What have we gotta do to get one of these script-minded artists around here to whip up a little Distance Between Particle plugin? Or I wish someone would get HVdeform working right.

These examples use a smoke plugin in Fusion written about 11 years ago. I love it for very fast-rendering wispy smoke, but is limited and has never been updated. But a simple DBP parameter would allow more similarly tendrily smoke and fire FX.

http://www.box.net/shared/static/28a5t0in6x.mov
http://www.box.net/shared/static/x4fpfq70nm.mov
http://www.box.net/shared/static/urr75v70id.mov
http://www.box.net/shared/static/18q4fhe1eg.mov

[DLIT]

Here are some tutorial setups for explosions with scenes and lwos but I can t interprete them once loaded in Layout. However some might find it useful and explain more.
Dom

[ridasaleeb2]

Mr Rid. Hi my name is Rida. Check out my "genie smoke" thread in "general" forum- love to hear from you on this thread. Also I find sometimes that particles don't mix together - rather show overlapping edges in a fire yellow or orange even with particle blur and crowd on. Rida