PDA

View Full Version : Splines vs. Polys


Hookflash
01-27-2003, 02:40 AM
Subject says it all;)

Hookflash
01-27-2003, 04:13 AM
Wow, splines are in the lead! I didn't expect that, even though this is the A:M forum.

Nurb'd
01-27-2003, 04:34 AM
post it in the general forum ;)

gra4mac
01-27-2003, 05:30 AM
A few days ago, I would have said splines no problem, but I have been playing with Wings and I find subDs very interesting.

Cheers, Graham

jaymackey
01-27-2003, 08:34 AM
I'm not sure that A:M users realize how much cool stuff could be done with polys, but that they will never see in A:M because of the hardline stance Hash takes against polys. My understanding is that all of the cool hair plug-ins are done with polys. A:M users want the popular hair plug-ins to come to A:M, but how can they port their plug-ins to an app that doesn't really have polys. There are the .3ds prop models, but are there Plug-in SDK commands that allow the creation and manipulation of polys?

Some users want fluid dynamics, so water, oil, goo, and other fluids can be simulated in A:M. There are plug-ins for other apps that do this, but they would have to be programmed from scratch to work with splines. People complain about Hash trying to do all the programming themselves, but they have to, because all of the mainstream plug-in companies write for polygon applications and not spline applications.

I don't think it's theoretically impossible to do these things with splines, but the API may not offer the ability to do it at this time. And the relatively small user base of professional users that would be likely to pay for such a plug-in versus the amount of work required to program it to use splines makes it unlikely that anyone will try it, even if the API is ready.

Spline do have advantages, but they have clear disadvantages as well. I'm not sure that adding polys to A:M is really the way to go though. My biggest fear is that Hash is simply not up to it. These are guys that have been programming for a long time, and should be very proficient. However, they may? have had more complete rewrites of the software than any other 3D animation App on the market. They're certainly in the top two. Yet it is a bug fest for a year after the last rewrite. That's not necessarily any worse than anyone else, in a sense: unlike Hash, the other companies would be working on the software in house over the course of that year, working out the bugs, so that customers would receive a working application, instead of getting it right away and being guinea pig-beta-testers for a year. The point is that they just may not be that proficient. Adding to the complexity of the program by supporting both splines and polygons, and allowing them to coexist and be used together, would probably be a recipe for disaster.

As much as I would like to see the latest polygon technologies added to A:M, along with a plug-in API that would allow a large number of popular plug-ins to be ported to A:M, their best bet might be to take Zpider's lead and start actually trying to make Splines easier and faster to work with, and to make sure that it is possible to write hair, fluid dynamics, etc. plug-ins for A:M. The second thing that they have to do is make sure that A:M is stable enough and capable of being used in a production environment. If they want people to write serious plug-ins for A:M, they have to make sure that there is enough of a user base that is willing to pay serious dough for them. Some of these plug-ins cost more than A:M.

My two cents... Maybe I'm off-base: let me know,

pabitel
01-27-2003, 11:36 AM
The reason third-party developers are not porting their plugins to A:M has more to with marketability. Most of the hair plugins, for example, cost 2x-3x more than A:M itself. Who's gonna buy it?

JoeCosman
01-27-2003, 06:12 PM
I know it may surprise many of you. but I'm gonna go with polys.

many of us remember working with polygons back in the 80s and early 90s. Those massive datasets, the different deformers, lattice cages, and effectors we all used to get the shape we want.
Using a billion points to make a curve. Oh yes, and every one of us stared in awe at the Viewpoint Dataset Catalogue. What heroes, to wrestle an hundred-thousand polygon skeleton.

That was a time when Martin's Splines Vs polygons essay really made sense. I mean, look at the alternative, back then! I think many of us in the poly world felt burnt about that era; lots of work for minimal functionality. It's kind of like one's very first car, which was a rust queen, and drank oil like water, and backfired, and scraped the ground going around turns... ahh, those were the days. :)

That was then. This is now. And these aren't your grandmum's polygons.

Getting from splines to polygons is fairly easy. The AM renderer is polygonal based. just dont let their mantra fool you. It should be changed to "just say no to modeling with polygons"

Thanks to programmers like Ed Catmull, Jim Clark, and countless others, we can do things with polygons to make them behave just like splines.
yup, you guessed it. Sub-D surfaces.

Sub-D surfaces are handled by the computer. No million poly dataset, we just work with our 300 polygon character and let the CPU do that stuff. Better programming, and modeling techniques have allowed manipulation of these surfaces to be pretty close to how splines work, such as tweaking the actual SubD surface (like in lightwave) and selecting a ring of continuous edges(the edge loop in wings). SubD's are actually simpler to understand in execution, in my opinion.

You want a bigger bulge in a bicep? Drag the four CPs that make up the curve away from the surface. No need to add extra CPs, unless you need a shape change. (yes, this is exactly the same with Splines.)

You need Gamma and magnitude? rotate and scale Edges instead of CPs. Need alpha and gamma adjustments? Rotate Faces instead of Edges.

Need hooks for extra detail in a low rez mesh? Triangulate the adjoining faces. need 5 pointers? Cap it with a 5(some programs like 4s and 3s instead)

SDS's not as lean as hash patches, you say? Since it's based on polygons, the 3 dimensional shape is defined by it's smallest components: triangles. Yes, you can make a spheroid out of 3 triangular faces with SubDs.

many programs will let you render perfectly smooth SubD's(such as Messiah) at any resolution. Many programs let you select a resolution to tesselate to (such as Lightwave and Maya), allowing for faster render times, and savings on RAM usage.

Some programs even let you piggyback CPs on top of other CPs(heirarchical subdivisions.) so you can increase detail without having to constrain them to another bone to smooth out a bend.

Polygons nowadays are very interchangeable between programs. Groups, UV coordinates, Color groups, quads, tris, normals. Moving models between apps has never been easier, thanks to open standards.

...Not to mention the excellent Free modeling programs that have sprung up recently.

GPU manufacturers are placing more emphasis on Geometry setup and tesselation inside the Video card instead of the CPU, making SubD and very dense polygonal models run extremely fast.

Polygons dont care about direction unless you make Games. You can have a dataset of tri-strips, or a globular mass of confusing triangles and it would still render smooth. creasing only happens where you want it to.

So... life is good for the polygon. it aint going away, it's just changing it's uses.

Use what you know. Know what to do. Do what you like. Like what you've made.

-Joe

Dearmad
01-27-2003, 11:13 PM
I think the bottom line is being overlooked here:

The crew at Hash probably couldn't do the intensive amount of programming required to add in anything other than the splines they already have. Look at the program's vector for the last versions (from 7 to 10). What has changed that is more than hype? let's see:

They added volumetric lights, motion blur (not really), particles, cloth, smart-skinning, spriticles, weighted bones, reorganized the workflow, what else? A couple of minor tools to further help with CP manipultion/implementation. Oh! Booleans (well sorta)! -snicker-

Does anyone see any *real* difference that is more than trimming or beyond basic programming? Ok, *maybe* the weighted CP bones will amount to something, but mathematically that's a pretty simple set of rules to place into the program, and so are the rules of particles if you don't need them to do much physical simulation with any degree of accuracy (and clearly Hash doesn't)- we're talking decades (50's) old math that was in computer raytracers available to the masses since the 80's with enough publicly available documentation to fill a warehouse (try every optical specialist physicists doctoral thesis since 1953).

Now, the latest developments in Sub-d and other manipulation tools for polygons are *not* that old. Why? because there wasn't the horsepower to even simulate doing this stuff more than twenty years ago (and that's about the lead time hash has shown itself to need) and the realm of development for them has been driven almost solely by practical application use- which means you need to be able to execute the theory.

And you want Hash to jump on this passing Porsche? They can't even get the interface to calm down.:wip:

The memory use of Hash's splines is minimal, as in insignificant. It's an interpolation process that (when computers were weaker) was brilliant.:shrug:

What was the question?

:shrug:

Natess44
01-27-2003, 11:51 PM
I like splines and I don't see the need to go over to sub d surfaces. If Hash can fix the problem with crease without having the user to tweek bias handlers all the time splines will be great. Hash should also put a subdivision type tool that adds splines though(It'd make life easier).

Roger Eberhart
01-28-2003, 12:05 AM
Originally posted by Dearmad
The crew at Hash probably couldn't do the intensive amount of programming required to add in anything other than the splines they already have.

Yet, Wings3D was programmed by one friggin guy and it's already three times the modeler AM ever was. And you know what, it didn't take him that long either.

I tried explaining to Ken Baer at a trade show why SubD's were easier to model with. It just didn't compute with him. The Hash guys are so convinced that their way is best.

It's pretty funny when a programmer tries to tell an artist what tools are best for making art with. I work at a game company. I know what 'programmer art' looks like. Thats the stuff they stick in until an artist gets around to fixing it. Boy is it ugly. Reminds me of Hash patches.

Nurb'd
01-28-2003, 02:30 AM
Yeah Roger but all he did was make a carbon copy of the modeling in Mirai and smaller Nendo. If anyone can say Wings is not a almost exact copy... Do not get me wrong here, he did a grat job keeping something alive. That is a very easy to use (free) modeler that works great. Still call it what it is.. a copy of a program many people loved.

Point still taken though... how hard would it be for Hash to do the same? The big companies steal from each other all the time. 3dsMax is a master at taking stuff from Maya/XSI and emulating it. Nothing wrong with that. When it is clear the r + d of hash is so weak they can not manage to stop crashes on 90% you have to wonder if they even could begin to add anything.

http://www.izware.com/

http://www.izware.com/nendo/product/modeling.html


The whole industry has switched to pouring resources into subd's so I do not think for a second believe they think there idea is better. I honestly think they do not have the resources to begin to put it in. What can they say but "our way is the best?" in wich I would reply "and that is why every other company has moved to sub-d right?" ... I mean ...

Sure splines are useful. Nurbs are still useful and that is why most packages have all three. Lightwave has simulated subd's and no nurbs and look at how succesful they have been. I can not see anyone using sub-d modeling for awhile and prefer splines. The finished product looks MUCH better.

The only way I would ever pick up AM again is if hash got serious and put it some modern modeling tools. If they did that.. then I would pick up a copy if only to support them. This stance that splines is better is nothing but a smokescreen though.

torpro
01-29-2003, 01:20 AM
I still like AM splines for the fact that you are actually creating and manipulating on the actual surface, not manipulating a cage as such. The other thing is the potential for the ArcticPigs plugin in conjunction with those same splines, Swf 3D models still take a time to download compared, then there is the materials, textures, transparancy etc that you get with full animated choreographies.

Hookflash
01-29-2003, 04:23 PM
Splines are getting trounced (well, sorta)! Long live polys!

*ducks and slinks off*

Natess44
01-29-2003, 05:57 PM
Whether you like splines or not I don't think Hash is going to get rid of them. So if you want to use polys then use wings or another program and import the model into A:M.

Hookflash
01-29-2003, 06:04 PM
Originally posted by Natess44
Whether you like splines or not I don't think Hash is going to get rid of them. So if you want to use polys then use wings or another program and import the model into A:M.

... and then spend 4 hours tweaking it to get rid of the artifacts.

JoeCosman
01-29-2003, 07:22 PM
Originally posted by Dearmad

Now, the latest developments in Sub-d and other manipulation tools for polygons are *not* that old. Why? because there wasn't the horsepower to even simulate doing this stuff more than twenty years ago


actually, Catmull-Clark surfaces were out in the late 70s. bezier curves were out in the 60s. there's quite a comprehensive article in a past issue of Computer Graphics World discussing the evolution of Subdivision surfaces(the one with Geri's game on the cover)

just because the horsepower wasnt there doesn't mean that those crazy ideas weren't ever around.

but, I do agree. Hash patches aren't going to leave AM any time soon. I've done some pretty nice models in AM, but after working with SubDs for over a year, my perception has changed. I open up AM and I can see every patch in those models.

it's not that splines are bad, its the math and method of getting them to the renderer that makes all the creasing. It wouldnt hurt to have martin take a look at sub-Ds to find a way to make his patches pretty again.

eboy
04-04-2003, 12:30 PM
Now if we could just get that wings3d guy to knock up an AM clone with added sub-d i'de be a happy man...

SouthernLaw
04-04-2003, 02:31 PM
For the time being, I'm gonna stick with Splines. Mainly because that's what I'm most comfortable with.

However... I've been playing with Wings 3D a lot and I really like the 'feel' of sub-ds. Modeling with splines is more akin to drawing in 3D whereas sub-ds seem more like sculpting. I've done *way* more drawing in my lifetime than sculpting - perhaps that's why I like splines.

Currently I'm using Wings to create rough models and then drop them into Hash for tweaking (sometimes a lot of tweaking) and adding detail. I find it's easier for me to get the basic masses and proportions of a character than in Hash. But I'm still not familiar enough to get all those little details to show up... everything kinda smooths out when I render.

daft
04-04-2003, 04:09 PM
Just a question? Those that are just voting for splines, have you actually tried poly modeling? Then I meen, really tried it. (*a voice comes out from nowhere* Maybe it´s just steve trying to balance out the over winning poly´s:)

I personally like both. Hash Spline are very good for animating and that is really only thing that is good about them.
Polygons are much better for modeling and I would like to see Hash make a better support for them.

koon69
04-04-2003, 09:21 PM
Yep sure have. Wings 3d - LW - C4d and Max. C4d the easiest "to get" and learn. Max is cool to LW is to weird for me Wings is great! I still like AM more! :> Feels more organic and natural to me.

Primus
04-05-2003, 01:09 AM
Subpatch (SUBD Poly) surfaces in Lightwave allows you to make models as complicated or as simple as you like. Creating a complex character with all the trimmings can take about six hours. I agree with JoeCosman. Years ago it was easier to make organic models using Hash patches, or at that time I was using Maya 2.0, NURBs. But with the ease and sophistication of SUBD surfaces nowadays...

(No Lightwave doesn't have true SUBD surfaces but it can still kick Softimage's and Maya's ass three ways to Sunday. It's nice to say "we have TRUE SUBD!", but in practice it really doesn't matter. Wings comes close but for me the lack of realtime cage manipulation is a down side.)

I've used Lightwave professionally for two years, I must say that A:M's animation capabilities is quite refreshing, really quite the opposite of its modeling. :) (I'm aware of people carping on about Lightwave's character animation flaws but it's quite usable, Messiah is just filler for me, it's harder but it's not impossible. People like me who come from Lightwave to A:M find the animation aspect pleasing, people who were weaned on A:M THEN shift to Lightwave have a completely different view, heh heh, you know who you are! Love that Lightwave baby!)

side note: A:M's ****ed up implementation of the timeline (what's the deal with 10 not being able to tack channels?!??) is frustrating though. Hell Lightwave's favorite channel set works better than 10's implementation!!!!

I've been using A:M it for a really involved project and I'm quite pleased with it. The project uses cartoony models with lots of character and lip synch, rendering is slow but the scenes are quite simple. Modeling is (excuse me...) bloody painful, more so than using NURBs, which is actually quite easier in organics (trim, blend, fillet...). But animating and setup is a... joy... :D

I tend to drift sorry, to say it succintly, splines suck. Polys/subd rules. No this isn't a matter of taste. It's just so. Anyone who says different can try convincing me with examples of models which are superior to something modeled in SUBD/Poly versus Hash Patches (BWAHAHAHA). (Sorry couldn't resist.)

Kalimol
04-05-2003, 04:51 AM
In the most utilitarian terms, it's a pain in the (censored) to have to restring splines and hook hooks and select and activate five-point patches to add the simplest little details, and polygon stuff has a lot of tools to work in details in much more straightforward ways, but I don't like the feeling of not working directly with the model, I don't like the increased potential to permanently ruin a good model (as splines are so much easier to rework and resculpt) and I don't like the computing power and disk space polygon models take.

I haven't used SubD--unless that's what I was trying to use in Wings :surprised --or nurbs, though, and I'm sure they're vastly superior.

Still, I can't say I'm thrilled about the idea of AM becoming one of the "big" apps to keep up with plugins. There needs to be an entry-level, hobby app out there, and Poser isn't good enough. (Not until they, y'know, remember to chuck in a modeling mode, thank you....)

binder3d
04-05-2003, 11:49 AM
I dont know - if I had LW I wouldnt waste my time with AM. Why bother with forums and AM? I would be in LW working away and not worrying over AM at all. I would be at the LW forum and hanging with the big boys. Yeah!
:rolleyes:

antonyw
04-05-2003, 07:13 PM
Originally posted by Hookflash
... and then spend 4 hours tweaking it to get rid of the artifacts.

I dont think so, u never can fix all the artifacts when u need closeup render.

A:M spline 90% same as version 3.5 till now, stilll have aritifacts/crease problems. Maybe they CANT FIX it because they dont know how to fix.

Why i still use A:M? becasue the Workflow really good/fastest.
1. Anytime can edit the model after added bone(maya/max cant).
2. Anytime add the Pose(blend shape) after added bone(maya/max cant).
3. Easy to bind(skin) character if you DONT NEED too accuracy blending.
4. Make/Modify 2d game(diablo like) character super fast because no need complex bone system and no need to care all artifacts.

hoochoochoochoo
04-07-2003, 08:44 AM
Lightwave doesn't have true SUBD surfaces but it can still kick Softimage's and Maya's ass three ways to Sunday.
Sorry, did I read you right? Are you talking about XSI or Soft3D (Soft 3D didn't have sub-d's) this is completely OT but I would disagree with you unless you've spent time with XSI. Well it is a public forum and I'm allowed to disagree...:beer:

Primus
04-07-2003, 07:23 PM
Sorry, did I read you right? Are you talking about XSI or Soft3D (Soft 3D didn't have sub-d's) this is completely OT but I would disagree with you unless you've spent time with XSI.

Yup :) I've tried out XSI. Great piece of software but the price tag really socks it to you. I still believe that Lightwave's modeler kicks XSI hiney when modeling organics along with Max, and Maya. Again Wings3D comes close IMHO, but without that real time SUBD cage manipulation it just misses the mark. For me it boils down to LW, XSI, Max, & Maya in that order. I model organics 100% of the time. I've tried out Cinema4D's demo, but never really had the chance to truly test it. A:M isn't even on the short list. (er for obvious reasons.)

(No, the lack of different Subdivision algos doesn't slow down Lightwave (XSI has CC and Doo-Sabin), the lack of true edge tools doesn't slow Lightwave down, Heck even Lightwave's inability to turn on subpatch when there's an N-sided poly doesn't slow it down, makes the mesh cleaner and the SUBD surface more predictable. I only use a handful of tools in Lightwave and what do you know, it works. Have you spent time in Lightwave? :hmm:)

Then again maybe I'm biased. :beer: eh? Cheers to agreeing to disagree!

binder3d
04-07-2003, 09:10 PM
Please dont take this wrong way - Im trying to understand better. If you have LW and its modeler kicks the c@$p out of XSI and Maya (no easy thing) why bother with AM at all? I know people say AM can be so much. But since its not and LW is already there why not just concentrate more on LW? Esecially with Messiah and all. Jut trying to figure this out. Thanks!

Primus
04-08-2003, 03:49 AM
Please dont take this wrong way - Im trying to understand better. If you have LW and its modeler kicks the c@$p out of XSI and Maya (no easy thing) why bother with AM at all? I know people say AM can be so much. But since its not and LW is already there why not just concentrate more on LW? Esecially with Messiah and all. Jut trying to figure this out. Thanks!

I don't need to concentrate on Lightwave, been using it at a company I worked for. That was my day job. I quit that job in order to set up my own small production studio. Lightwave is an excellent piece of software. I wouldn't hesitate to sick it on to anyone. But economic realities will cast its ugly head. I've used Project:Messiah in the past, but mostly pure Lightwave.

The short version binder3d is this: Big project, needs ton of animation using toon characters for educational purposes with lots of voice and lip synch. I set up a studio with 4 animators with myself as TD. I needed a package that can handle the above on the cheap. Forking over $5,000 dollars (with shipping to my er SARS infected part of the world) for four seats of Lightwave was not a viable option, since I was financing the whole four month gig. So I resigned from my company, gave the standard 15 day notice, (who BTW still keeps me employed as a part timer...) got some advice from lots of forums, and finally purchased four seats of A:M, Net render and a ton of educational CDs and the A:M 2002 complete guide. Total cost was $2,000 plus shipping. Here's an earlier thread [URL=http://www.cgtalk.com/showthread.php?s=&threadid=45873[/URL] giving the exact details.

I'm happy to say with the exception of modeling and rendering A:M is keeping up with our schedule. Coming from another software package, I'm happy to say that knowledge is transferable and we were up to speed in A:M (with help from those educational materials, we bought one of the CDs with an educational discount and got all that stuff for a hundred bucks more! Shipping costed more!) quite quickly. It really is an excellent and very affordable package. 10x is kinda stable. It doesn't crash on us. It just... dissapears... hmmm strange.

With the new version of Messiah, it would cost more, about $6,500, with four seats of Messiah:Animate, and one seat of Lightwave for rendering using screamer net.

A:M also has that handy automatic dope sheet, kinda like Magpie Pro, which was pretty tedious in Lightwave. Since 90 percent of our animation revolves around a medium shot and a talking head... :D But... we still needed to be faster, so we're looking at mixing a lot of 2D animation to our pipeline (we all come from a traditional background, two of us came from Toei doing in-betweening), we're looking at Moho which costs about 90 bucks. Virtually no rendering time. We have 5 machines and Net Render at four minutes a frame it's still painful. Clear? :cool:

binder3d
04-08-2003, 04:18 AM
Cool - thanks! Have you tried 10.5? From what I have seen the render is way better and getting better all the time. I give you major props for quitting your job!
:buttrock:

Primus
04-12-2003, 03:21 AM
No problem, they've kept me on part-time though (my day job) :) I downloaded the latest 10.5 Alpha. Some of the features that were supposed to work, now kinda work. Five point patches get error renders. The push pin now works (kinda) in the timeline window. Very strange.

We're using the latest stable download, and it's pretty stable :D Great, great piece of software. Very quirky modeling, but... It gets the job done.

proton
04-12-2003, 06:20 PM
Use what works for you....That's what I always have said.....Take Raf Anzovin for example....he uses them all....depends on the job......Let's bring the Software bashing and comparisons to an end....Artists should use what works for them....

I'm excited about bringing AM back into my toolkit....in other words...


Can't we all just get along:love:

binder3d
04-12-2003, 08:56 PM
Thank you! Thank you!Thank you!Thank you!Thank you!Thank you!Thank you!Thank you!Thank you!Thank you!

I agree Proton. I think its time to move away from hey this app is good - but it sux to. One good thing and three bad is not very constructive!

Proton-do you use Wings 3d to model or LW?

proton
04-13-2003, 04:40 AM
I model in lw

John Keates
04-20-2003, 01:34 PM
Originally posted by binder3d
I think its time to move away from hey this app is good - but it sux to. One good thing and three bad is not very constructive!


What counts as constructive then? Why can't we complain that we don't have splines AND subds in AM?, most programs do.

binder3d
04-21-2003, 09:50 PM
I think having subdiv in AM would be fantastic!!!! But it seems to much that posts in general seems to hack at AM just for being AM. Its one thing to say subdiv would be great-which I agree-but to say AM needs this and that becasue AM suxs is not very good for AM. I like constructive criticism if its not mean and unwarranted. Thats all! :>

iBlue
04-30-2003, 08:01 AM
nothing like subpatch in LW! without it, i dunno HOW i would model.. it would prolly be .. weird o_O

CGTalk Moderation
01-14-2006, 06:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.