PDA

View Full Version : Is hardfx actually usable?



alexos
10-27-2008, 04:10 AM
I mean, yes, I did break down the customary brick wall with the medieval-looking demolition ball. Many times. And the vase, and the glass, and I sent a car through a glass panel - it worked.

But last week-end I decided to try and do something a tad less chrome-ball-on-checkered-floor-ish, and went for an actual - you know - scene: a typical, everyday situation with a meteor hitting a building and subsequently the road below, scattering an amount of innocent cars around in the process.

Well... turns out I couldn't. I mean, apart from the obvious problems (me struggling to figure out the setup being the most prominent) the calculations just went on forever - and I mean forever as in, 30 minutes at frame 1 with no visible clue on how long it would take.

Now I always assume I'm doing something wrong and/or missing The Trick that will speed things up, so I'm hoping someone will point me to that; but meanwhile, I wonder...

ADP.

Surrealist.
10-27-2008, 04:51 AM
With more objects and vertexes it is going to take a lot longer to calculate. That is one thing. Something that will help you with the scene scene is you can think about what things you can be faking, that is, keyframing or any other method, and what things to do with dynamics and what things to use FX hard link on from a low poly proxi object. Then of course you may also realize you need to do the scene in several passes for each effect. (as an example low poly proxi objects for collision objects to calculate the collision, save out the motion and use that or render it with alpha masks) I don't think it is possible yet to just set up a large scene like you do in a real movie and then call out action (hit calculate) and watch it unfold. For computer animation we still have have to think in stages and break things up into various parts or effects. Computers are not fast enough to do everything yet. So think about what you can fake or what you can break down into parts. Cheating is everything in real-world filmmaking and computer animation is no exception. Hope that helps to put things in perspective and give you some ideas. :)

EDIT: As an example recently I had to do a large amount of geometry for a set. Turns out realistically I could only render out about 1 12th of the whole thing at a time due to ram limitations. So I had something like 12 or more layers in my NLE that I was compositing just to make one frame!

Dirk
10-27-2008, 04:55 AM
This usually happens to me if some bounce values or forces are set too high. I think HardFX goes into an endless loop then.

3DGFXStudios
10-27-2008, 05:31 AM
I just think hardfx just sucks....I mean all the lw dynamics suck. Just look at the new dynamics in game engines.

Dodgy
10-27-2008, 07:28 AM
How many polys are in the cars that you're throwing around? As the number of polys goes up,the calculation time goes up exponentially. Are you using any proxy objects? Are you using the Box or Sphere collision types? There are any number of ways of speeding up calculations before you say 'Are dynamics usable?'.

That is how game engines do it, by a lot of short cuts and hacks, which they can get away with because the player isn't looking too hard. I know one of my programmer friends fought a long battle trying to get unreal's dynamics to not suck. I'm not saying LW's dynamics are state of the art, they are a few years old now, but I've seen a lot of 'LW's Dynamics suck' threads and a lot of the time it was lack of knowledge which caused the 'suck', not the capability of LW's dynamics.

Would you care to share a scene for help?

SplineGod
10-27-2008, 08:28 AM
Heres something lino did thats similar. He used LW.
http://www.youtube.com/watch?v=kOinjz10e5Q

alexos
10-27-2008, 09:49 AM
Heres something lino did thats similar. He used LW.
http://www.youtube.com/watch?v=kOinjz10e5Q

Yeah, that damn bridge. I'd seen it before - Lino, if you're reading this - I hate you :)

I have no clue how he did that, really. And Dodgy - far from me to say that LW dynamics suck. I have zero experience with dynamics (I work in architectural visualization, a field where for some reason you don't often get to destroy stuff) and I'm certainly not one of those "but you could do this in Maya in about 20 seconds!" types. All I can say is, my cars were fairly low-poly ones (about 2k polys each, and there only were four), the buildings were as low-res as I could make them and the road, well... was straight. I think the entire scene amounted to, oh, 70k polys or so.

I tried sphere and box, yes - to no avail. As I said, I was just wondering... does it -for instance- calculate the *entire* scene dynamics? I mean, if my meteor (a mere null object, by the way) hits the building first, the road next and the cars last, does it try to figure it all out at once or does it take it step by step - building first, etc.?

I will try to split the scene -which I can't share, for I didn't bother to save it :)- in smaller pieces and see what happens... and I'll check that I don't have any unreasonably high values. Meanwhile, thanks.

ADP.

3DGFXStudios
10-27-2008, 10:35 AM
Heres something lino did thats similar. He used LW.
http://www.youtube.com/watch?v=kOinjz10e5Q

that's actually quite impressive! But how did he do that? :stumped:

shrox
10-27-2008, 10:45 AM
that's actually quite impressive! But how did he do that? :stumped:

There is a brief wireframe render:

http://www.youtube.com/watch?v=oIscv7mfQ04&NR=1

prometheus
10-27-2008, 02:50 PM
that bridge collapsing is nice..

but I donīt think thereīs any real selfcollisions in there and That is the one main thing that hardfx canīt do well..together with lots of cracked pieces it can slow down. otherwise I wouldnīt say hardfx sucks..it requires some late nights and test all settings and stay alert to tutorials and tips and tricks around here.
what makes that bridge colappsing good ..I would say the overall look, rendering, camera, and timing.

Michael

hruffin3
10-27-2008, 03:38 PM
setting visicosity high really taxes you - calculation takes a lot longer.

Sande
10-27-2008, 05:13 PM
that bridge collapsing is nice..

but I donīt think thereīs any real selfcollisions in there and That is the one main thing that hardfx canīt do well..together with lots of cracked pieces it can slow down.
Of course, you usually don't need self collisions on all pieces. Especially when you're breaking stuff with lots of debris, dust etc. Actually, it is one of the key optimizations to use those collisions just with the large, obvious, pieces that you can't get away with cheating and have a different set(s) of smaller pieces without self collisions.

One of the things that also make the bridge scene look so good is the nice use of hypervoxels - they can really, really, boost the scene. If you look at the wireframe the dynamics aren't that special, but when coupled with nice lighting, motion blur, dust and high velocity debris flying by, it all starts to look amazing...

Most of the time the less you actually see of the action, the better it looks. :)

Mr Rid
10-27-2008, 06:27 PM
I mean, yes, I did break down the customary brick wall with the medieval-looking demolition ball. Many times. And the vase, and the glass, and I sent a car through a glass panel - it worked.

But last week-end I decided to try and do something a tad less chrome-ball-on-checkered-floor-ish, and went for an actual - you know - scene: a typical, everyday situation with a meteor hitting a building and subsequently the road below, scattering an amount of innocent cars around in the process.

I would say LW dynamics are very limited compared to dynamcs in almost any other professional app claiming to have dynamics.

But you stumbled into an important point I like to keep making about whimsical demos as proof of how capable an app is. It is one thing to do what many do and just play around with simple tests with spheres and cubes or big chunky, angular pieces. But it is an entirely different animal when you get into a more specific scene with complex realism. The control in LW tends to be ever frustratingly lacking for a production environment. You usually have to do a lot of McGuyver workarounding.

Lino's bridge is impressive at first glance, and I recall that he put a lot of time into this and several iterations before getting it to look decent, although still a ways off from photoreal. Was hardly 'a snap' in LW dynamics. But the hardFX pieces are large, chunky and sharply angular and do not continue on to collide with anything. And it is usually MUCH easier to do even something like this when it is just for your own whim, rather than having to work within constraints of a practical shot or picky client.

I just wish NT would follow thru with more competitively advanced control. In general, LW features tend to be good ideas that too often just peter out at 'minimally functioning' and then dont get significantly upgraded for a number of versions.

But what you are describing will take some real McGuyvering or layering of multiple techniques and workarounds along the lines of advice already given. It is correct you wont be able to just turn HardFX on a complex scene and watch the show.

prometheus
10-27-2008, 06:59 PM
Lino's bridge is impressive at first glance, and I recall that he put a lot of time into this and several iterations before getting it to look decent, although still a ways off from photoreal. Was hardly 'a snap' in LW dynamics. But the hardFX pieces are large, chunky and sharply angular and do not continue on to collide with anything. And it is usually MUCH easier to do even something like this when it is just for your own whim, rather than having to work within constraints of a practical shot or picky client. .

yah..I guess the true control ends when your not doing an fx war shot, and when it is needed for a client I guess.

The client might go..now I want those pieces over there that already are exploded high in the air to collide with something else and at the same time be fragmented on impact...now, how do we do that with workarounds..huh.

Michael

prometheus
10-28-2008, 07:54 AM
Well one thing thou..for me crackin up an object and letting it smash to the ground and break is one thing and itīs very fast and easy to do in Lightwave..Im having some hairloss acheiving this in Houdini, still very newbie of it.

hereīs a thread trying to explain why things donīt break apart so easy in houdini..Im not sure if you need to be registred to read this thou.

http://www.sidefx.com/index.php?option=com_forum&Itemid=172&page=viewtopic&t=13205&highlight=rbd+fracture

Iīm having a learning threshold to get by to be able to understand all the sops and dops and flags and nodes and connecting it all, I do understand that thereīs a lot of power in this nodal procedural structure and at the end there are things possible that just canīt be done in lightwave I guess, but as an " artist" I am not that fond over the complexity in doing this, I surely hope that when an improvement to hardfx is done, and if they implement nodes, I would be happy if it stays with some simplicity.

Michael

heynewt
10-28-2008, 09:04 AM
Like everything in 3D, the manufacturers give you tutorials to lead you to believe it's just push a button and 5 minutes later your hardfx object explodes into a billion perfect pieces. In the real world it doesn't work that way. Those are serious simulation calculations going on.

It takes time and it takes experience to get hardfx to do spectacular looking work. One of the main "tricks" (as was stated above) is to take your object and divide it up into many small objects. Then do calculations on each one.

I just blew up a convenience store for a client 2 weeks ago. I had to divide the store into 9 separate objects, and it still took me an incredibly boring 9 hours of sitting here watching each section calculate its hardfx then saving the .mdd file for that section. Move onto the next section, calculate its explosion, etc. The bridge that lino did was done one small section at a time. Doing cool big explosions takes a lot of patience and practice doing small explosions. If you're computer is choking, then divide your objects up into smaller sections and do one at a time.

Who knows what OTHER big trick in Lino's bridge collapse really sold it? Had nothing to do with LW. Anyone? Buehler?

Sound. Watch it without sound and with sound. it's still impressive, but sound design adds TONS to selling the visual nature of an explosion.

Robert N.

prometheus
10-28-2008, 01:39 PM
Sound. Watch it without sound and with sound. it's still impressive, but sound design adds TONS to selling the visual nature of an explosion.

Robert N.

yepp thought so too..The sound is really important to sell something..
Makes me think of some extrem Hard fx selfcollision impact sound every time a piece is colliding with something, now..how do we go about that..huh, nahh!

Instead of sound editing with fakes in post, why not pull real impact data out of simulated hardfx impacts that could be translated to sound effects...ehh..perhaps letīs start by getting true hardfx self collsions and some efficient masscollision calculations.

Michael

Surrealist.
10-28-2008, 03:12 PM
Yeah the best solutions are usually a combination of things that each don't have to carry all of the weight themselves. You kind of have to spread the task to a few things.

Somehow, it is easy to forget that this is filmmaking. It is he end shot that is important. It is what winds up on the screen that is going to be looked at. Does not matter how many pieces of gaffers tape are holding up the rafters in the background as long as it looks right on camera. Tilt the camera a little to far and oops, there's the guy holding a bag of "snowflakes" standing on the top of a ladder. :D

And also each shot you make or are making is a part of a sequence. Or well, it should be. A thought about how many shots and from what angles are going to take up your explosion sequence is just - if not more - important than the explosion itself. And changing a shot is a great way to get out of a problem or into a solution.

In one shot it might be better to key frame some action of cars being knocked around with some dynamics explosion as a background sequence or mapped onto a poly in the BG.

In another shot your focus might be on the actual first contact with the building where the impact hits and bricks fall with some explosion. In that shot alone you could have several passes just to get the effects just right. Not only that. It is much more efficient to work on one thing at a time and composite. This way you only work on the dynamics in one pass with only the essential things that have to be dynamics. In another pass of that same shot you might have say just the key framed elements such as cars in the BG or what have you. In another pass just the hypervoxels or other smoke effects. This is not just practicality this is also a smart and easy way to control each element.

So at no time should you really have all the polygons in th4e scene taking up ram and in general slowing things down. You can build a low poly proxi set to fill in for all the elements that are going to be there in the final shot. This is essential to visualizing your camera angle and what is going to be in that shot.

And finally, you can cheat anything. It is a great idea not to build your large set pieces as one object. Build them in parts you can move them around freely if you need to to get the composition just right. Does not matter where they "should" be it only matters where they look right or make the shot work compositionally just as long as you can sell the shot and it does not look wrong.

Mr Rid
10-29-2008, 11:14 PM
yah..I guess the true control ends when your not doing an fx war shot, and when it is needed for a client I guess.

The client might go..now I want those pieces over there that already are exploded high in the air to collide with something else and at the same time be fragmented on impact...now, how do we do that with workarounds..huh.

Michael

Yes, the client will ask, 'Can we have a lot more...' 'we decided to make the sequence 200 frames longer/shorter but keep the same action', '... is now at 120 fps', 'just move that part over/down/back, speed this part up and slow that part down, mix in more cars and bodies', 'no, no, no, more like in Matrix/Star Wars/Gandhi', 'that should happen later, more spread out, have these pieces go that way and those pieces shatter into smaller pieces, burst into flame, bounce off the dancing dolphins... what, didnt you know about the dolphins? Yes, well I had this great idea last night. Its all under water now.'

heynewt
10-30-2008, 06:25 AM
Holy crap, you and I have the SAME client! ;)


Yes, the client will ask, 'Can we have a lot more...' 'we decided to make the sequence 200 frames longer/shorter but keep the same action', '... is now at 120 fps', 'just move that part over/down/back, speed this part up and slow that part down, mix in more cars and bodies', 'no, no, no, more like in Matrix/Star Wars/Gandhi', 'that should happen later, more spread out, have these pieces go that way and those pieces shatter into smaller pieces, burst into flame, bounce off the dancing dolphins... what, didnt you know about the dolphins? Yes, well I had this great idea last night. Its all under water now.'

Mr Rid
10-31-2008, 01:08 AM
Holy crap, you and I have the SAME client! ;)

Dont we all.

I usually find involved LW dynamics/PFX setups to be like a house of cards. Once I've carefully finagled it into standing up by itself, slowly step away... no one breath... easy... there, its working. But then the client inevitably wants me to start moving cards around... :cry:

Stooch
10-31-2008, 03:10 AM
luckily i have realflow to offload some of that stuff. but yeah dynamics are probably lws weakest area at the moment.

they just added fur too, you woould think that dynamics would be at the top of the list.