View Full Version : Creasing and the future

 hoochoochoochoo02 February 2003, 01:34 PMThis has come up on some of the threads here (and will never survive on the main AM list) but out of curiosity and maybe Hash will pick it up. Creasing, if you can model around it (and patience or even moving onto another model then returning helps) then fine but what about when you are trying fairly low resolution work? XSI has Geometry approximation, I've heard of "meshsmooth" in other apps but never tried it - so what other ways could this be achieved? I finally got to play with porcelain.mat then realised it was too much of a blunt tool. Would a slider help? (so you could set the degree of smoothing you wanted?) I was looking at the Renault website and came across the regards to Mr Bezier - would Hash Splines still be as flexible if they had a "Bezier-spline" like ability and could that get rid of some (I know it won't get rid of all) of the creasing that sometimes happens? just some thoughts - shoot me down if you wish.
JoeCosman
02 February 2003, 06:33 PM
right now, Hash patches are based on an obscure technology called Coons patches.(I say obscure, because the only app I have ever seen it in is A:M, and freeware A:M clones)

While Martin and co. are adamant about sticking with their patch technology, there wouldn't be any harm in using bezier math for patches instead.

Anyone who has worked with Strata Studiopro(the old one), sratavision 3d, Illustrator, photoshop, Quark, Clarisworks, and many drafting applications can understand why Beziers are the bomb. :-)

for one thing, a Bezier control point lies ON THE SURFACE.
The bias handles can move in any direction, and like hash patches, can be broken and moved independently of each other.

a bezier surface is ACCURATE. You can define a perfect circle with Bezier math.

Beziers are efficient. you can define a perfect sphere with 2 bezier patches, and a perfect circle with 2 bezier splines.

Bezier math is at the root of NURBS algorithm.(the Acronym NURBS means Non-Uniform-Bezier-Spline)

The algorithms in hash patches are....should I say... less than acceptible these days. The amount of responses to Mesh smoothness and creasing on the A:M list and here at CG talk is quite phenomenal.
If I were Hash(and I'm not) I would at least be entertaining these ideas, just to see what the outcome looks like, instead of poohpoohing suggestions left and right, and bathing in hubris.

John Carmack of id software does this all the time, producing custom builds of his game engines and code, just to see what works and what doesnt, what looks better, what goes faster. He acts on his ideas and finds out why they can't work, or why they do. no assumptions.

"the funny thing about ideas is.... they won't work unless you do"

Goon
02 February 2003, 06:37 PM
I don't know if a b-spline would have less creasing, from what i've read on these boards it has more to do with how the surface is handled.
Have you played with Bezier curves in Illustrator? Those are just about the coolest things ever.You can add points with changing the curve, change the magnitude, angle, break the handles, control the angle of each handle individualy. I don't know how well these would animate, but if they could be implemented that would be so much control.

JoeCosman
02 February 2003, 06:55 PM
NURBS surfaces are like this. Multiple nurbs patches joining into a corner, and yet... renders smooth.

the reason being, is that the tangencies of the joining patches are matched together. its a bit hard to explain. here's an A:M example.

you want to make a fillet(a curved transition from one mesh to another) for a wing. You make both objects, join them together and get really dissappointed at the junction.

you immediately peak the perimeter(making the bias handles independent) and start tweaking the gamma and magnitude.

you watch closely at the patch surface, staring at the peaked crease, adjusting the bias until it seems to *dissappear*

you do this same experiment with two meshes, their edges and CPs are lined up perfectly, except... its very lumpy. you immediately grab the bias handles on the edge of the first object and start rotating them until you see the edge dissappear.

now, theoretically speaking, if the curve looks smooth, then selecting both bias handles on both meshes should show them going in a straight line, as if they were made from a single mesh.

the same manual technique would work for Bezier spline patches. the big difference is, the spline curve transitions more smoothly to the next CP, where as, you will get an unnatural lump with a hash spline.

This is how multi-patch nurbs work, except the computer is keeping the surfaces tangent, instead of you.

the point is. the modeling paradigm wouldn't shift at all. Just the way the math is handled, and that seems to be the one thing they will never change. Sure, they'll code workarounds and render fixes(porcelain) but it will never change.

John Keates
02 February 2003, 01:17 PM
Originally posted by JoeCosman
the point is. the modeling paradigm wouldn't shift at all. Just the way the math is handled, and that seems to be the one thing they will never change. Sure, they'll code workarounds and render fixes(porcelain) but it will never change.

Why is this exactly? Is it shear arrogance or have any reasons ever been given? How hard would it realy be for them to implement a new type of spline? After all, most software has a multitude of the things.

Kalimol
02 February 2003, 01:02 AM
Well, first, it's a function that's at the core of the program--everything from modeling to rendering is built around Hash's splines. And looking at their help file's little party line explanation of why splines are the future, I think arrogance plays a key role as well.

Drakkheim
02 February 2003, 06:17 AM
Well This little discussion got me interested in some of the math and types of math involved. So i did a bit of looking around on the web and found the following pdf file which describes coon patches , bezier and other types of patches. just thought ya might like to fry a few brain cells.

-http://www.me.mtu.edu/~bettig/MEEM5990/notes22.pdf

-d

ypoissant
02 February 2003, 05:23 PM
One of the causes for creases on Hash surfaces is because the tangent at each CP is constrained to the two neighboor CP on the curve. It is not because of the type of curve. It would be possible to do that for any other types of curves.

Take any vector based program like Illustrator and draw a curve. Now move a CP from the curve on the screen. Notice how the tangents of the neighboor CPs stay oriented at the same angle independent of its relative position to the moving CP. If you where to use that in an animation, this would mean that each time you moved a CP for surface deformation, you would also have to tweak the adjacent CP tangents. That is because each tangents are independent from their neighboor CPs position. Each CP tangent angle is relative to the global space which is the sheet of paper.

Not so with Hash curves. Here each CP tangent is relative to a localized space of reference which are defined by the direct neighbooring CPs. Such that if you move a CP, the tangents for the two neigbooor CPs automatically moves too.

Given that A:M was designed primarily for animation, I think that this implementation was a brilliant idea. But unfortunately, it produces creases.

Relaxing the local space of reference is not a good idea IMO because this would mean that we would have to deal not only with moving CPs but also with adjusting neighboor tangents as well when animating. Sub-D seems to be more similar in their working because there are no tangents to tweak and that was the intent (I guess) when the Hash splines where designed.

Another web page about curves (http://www.cs.sfu.ca/~torsten/Teaching/Cmpt361/LectureNotes/HTML/08_2Dcurves/)

Yves

howardt
02 February 2003, 06:53 PM
Yves, I don't quite understand what you are saying here.

As you say, the 'type' of curve is not the issue here. In fact, Hash's curves are Bezier cubics, and the tricky part is, as you say, that the tangents are based on clever formulas involving adjacent CP tangents, to enable easier animation.

The weird thing about Hash patches is that the interior points are not defined by a formula based on the edge curves, but rather on a recursive procedure (subdividing along one dimension or the other, according to the "curviest" side). I don't know if someone has proved that this procedure is equivalent to Coons patches, but I kind of doubt it. Now, the crux is: bicubic bezier surfaces have 16 control points, but Hash only lets you play with the 12 on the boundary, leaving the interior 4 up to whatever its subdivision algorithm does. It would seem to me that to avoid creases, something has to be done to calculate those interior 4 control points so as to match tangencies across the edges to adjacent patches.

ypoissant
02 February 2003, 10:02 PM
Howard,

I'm glad you get into this discussion. I think you are one of those few with in-depth knowledge about Hash patches and their equivalency with other types of curves.

I was basically saying the same thing as you did but tried to use non technical words. I thought the "illustrator" example provided a practical example.

I did not wanted to get into how the patches themselves are derived but since you brought it up... As you said, the patches are computed through a subdivision scheme and not from those 4 additional CVs that we find in Bezier or NURB patches. But to me, the fact that it is derived from subdivision or explicit through those 4 additional CVs is not the issue as far as creases are concerned.

What really causes creases is the fact that the tangents are parallel to the two neighboor CPs by default and that even if you tweak them so that they are not parallel anymore and you have a better looking surface, you still have this local space of reference which may cause pops when animating.

By local space of reference, I mean this fact that the tangent is relative to this local space aligned along the 3 CPs that are the CP which gets its tangent tweaked and the 2 neighboor CPs (the next and the previous CP along the spline).

So Hash patches are subdivided but unlike Sub-D which don't have any explicit tangent that you can manipulate, Hash patches do have explicit tangents but then, unlike other type of patches which aren't constrained to the neighboring CPs, Hash tangents are constrained locally.

This constraining of the tangent to the neighboring CPs is by design and is not an implicit feature of the splines used. The same decision could easily have been made for any modeling system based on any other types of piecewise patches (the classes of patches where the curve meet the endpoints though).

So in order to reform the Hash patches, Hash would have to either :
a) find another local space of reference definition for the tangents which would behave in a more intuitive fashion for the modeler in order to keep the explicit tangents. Personally, I don't think this is possible. Or
b) abandon this local space of reference and thus leave the tangent orientation exclusively to the modeler/animator responsibility. I don't think users would like that. Or
c) adopt a pure subdivision scheme and abandon the explicit tangent altogether.

A tough decision for sure.

Yves

howardt
02 February 2003, 01:07 AM
Yves, I guess I understand what you are saying, but I don't think I agree. However, you have much more modeling experience than me, so you are probably the one that's right here.

Nevertheless, I'll try to explain my thoughts. I would think that creases are formed because, at the middles of the edges, the surface tangents perpendicular to the edge direction don't match up for the two patches that meet at the edge. It would seem to me that those tangents are most affected by those interior 4 CPs that I was harping on above. It is my feeling (that I can't back up with math right now) that you can have arbitrary tangents at the corner points - as long as the incoming and outgoing directions line up, as they do if not peaked - and that proper setting of those interior CPs will avoid creases. The smooth Bezier patches Joe was talking about above. That's why I don't buy that the constraints on the tangents at the corners is what is causing the creases.

EROMLIG
02 February 2003, 03:04 AM
I wouldn't mind seeing another form of modeling included even if it were limited to models that could be moved but not deformed. Just as only polys with 3 or 4 vertices can be sub-d'ed in Lightwave, but you still have the option to work in the standard polygon modeler or mix the two. It gives you the freedom to use the solution best suited to the task at hand. If hash splines aren't changed to behave very differently I don't see the program becoming that much more appealing to folks wanting to do much more than "cartoony" subject matter. It's not just the creases. The inability to add detail easily, the problem with splines that occur too close to one and other on a continuous surface. There are a number of things that IMHO the modeler just does not handle all that well. Honestly, Howards font/ai wizard was the first real additional tool to be introduced in the entire time I had been using the program.

ypoissant
02 February 2003, 04:45 AM
Howard,

I have a tutorial about gammas for A:M at:
http://www.ypoart.com/tutorials/gammas.htm
which explains the creases and how to avoid them. But here is a demonstration of what I mean:

http://www.ypoart.com/tutorials/pictures/gammas/gamma001.gif

See how the tangent for CP 3 is parallel to the space of reference aliigned on CP 2 and CP 4.
http://www.ypoart.com/tutorials/pictures/gammas/gamma002.gif

As a modeler, you would rather expect a tangent like this:
http://www.ypoart.com/tutorials/pictures/gammas/gamma003.gif

Now, back to our default tangent, if we extrude this spline we get something like that:
http://www.ypoart.com/tutorials/pictures/gammas/gamma015.gif

I think, from the above picture, it is easy to see where the crease would be visible.

Here is the extruded surface in choreography:
http://www.ypoart.com/tutorials/pictures/gammas/gamma-anim.gif

It is very obvious that the creases goes all along the spline. And if you tweaked the first spline before extruding (well, you could also tweak the tangent on the 3 splines), the crease would not be visible and you would have a smoothly shaded surface.

I hope this have made my understanding of the creases more easy to understand.

BTW, since at least v10, you can now tweak your tangent without fear of pops during animation. Tweaked tangent behave intuitively now in poses or actions. You can even animate the tangents in a pose.

Yves

howardt
02 February 2003, 01:03 PM
Thanks for the picture Yves. Now I understand the kind of creases you're talking about, and I agree with you. Hash's method for setting tangents "locally" (as you say) has produced many more points of inflection than you'd like, and it would be hard to dig out of that creasing hole. Maybe they could change their default tangent calculations so that cases like these work better: that is, avoid 3 points of inflection in a 3-cp-segment when 1 could be achieved.

[For those who only vaguely remember calculus: Point of inflection = place where curve transitions from "concave" to "convex"; mathematically, the second derivative changes sign]

There are other sources of creases in A:M. The one users hit first, I think, is when you have more than two splines crossing at one CP - the worst is when you have a good "X" pattern, and then bring in another spline to terminate at that point. This is where the old manuals used to tell you to leave an unrendered tail on the other side of that third spline, and adjust the tail until you get a smooth result. Yuck. It is this kind of case where my thoughts about the interior 4 CPs above were more directed.

ypoissant
02 February 2003, 04:11 PM
There are other sources of creases in A:M. The one users hit first, I think, is when you have more than two splines crossing at one CP - the worst is when you have a good "X" pattern, and then bring in another spline to terminate at that point. This is where the old manuals used to tell you to leave an unrendered tail on the other side of that third spline, and adjust the tail until you get a smooth result. Yuck. It is this kind of case where my thoughts about the interior 4 CPs above were more directed.

I would also add the ")(" pattern to the list of no-no.

Yes. I see what you mean. But the proper solution to these creases is to not do that. That is why hooks and 5pt patches have been added to A:M. Without hooks and 5pt patches, the situation you describe where simply unavoidable on any practical models. But since hooks and 5pt patches, I have yet to find a situation where the inelegant solution you mention are totally unavoidable. Well one of my beveled primitives don't have an elegant solution but I could choose an inelegant model that would not involve inelegant splineage crossings though.

Yves

3DArtZ
02 February 2003, 05:04 PM
"right now, Hash patches are based on an obscure technology called Coons patches.(I say obscure, because the only app I have ever seen it in is A:M, and freeware A:M clones)"

The electricimage modeler has a coons surfacing tool.... don't use that feature too much
but the smoothing features in EI modeler rock. I'm sure other nurbs modelers have it as well.
In my opinion the best way to get rid of your creasing at the moment is post production.
If you have photoshop.... you could batch process a folder that holds your individual animation frames.
Using Photoshops action feature(records step taken on file)...
copy original frame to new layer(layer1)
Guassian blur original frame(backgroundlayer), apply layer1 to background(blurred layer) with overlay...
set layer1 to darken which will excentuate the hard edges of your blurred image making it look sharp again while the content of the items in the image will remain smooth(er).
This is what I've been doing to get rid of creasing on my models... works for me, maybe for you too.
Once your record your steps taken on one frame, you can tell photoshop to automatically open, go through steps,
and save everyfile while you watch tv, or shovel 2 ft of snow from your drive way!
Mike Fitz

John Keates
02 February 2003, 07:58 PM
Originally posted by ypoissant
I would also add the ")(" pattern to the list of no-no.

Without hooks and 5pt patches, the situation you describe where simply unavoidable on any practical models.
Yves

I am not totaly happy with 5 point patches. Sure, I use them all the time and they are usefull, but it is hard to avoid making them look crickly or hollowed out. I know all the rules about how to use them and everything but I would love to be able to just slap them on without worryinging about sharp crinkles. I am sure that the folks at Hash can come up with something.

ypoissant
02 February 2003, 10:45 PM
Mike reply makes me think... Why not just tweak the damn tangent so that the crease go away definitely and thus avoid this whole post-processing step?

If you where using any other spline based patch technology, you would have to tweak the tangent as part of the modeling process. A:M facilitates the job in the sense that most of the time, you don't even have to tweak the tangent. But whenever a crease pops-up, you still have the option to correct it with the tangent handle adjustments. So why not just do that?

Just currious :shrug:

Yves

hoochoochoochoo
02 February 2003, 11:34 AM
Yves, you have stunned me with your information BUT I'm with Joe Cosman on this one. I respect both of you as modellers but I hesitate at having to go into heavy duty calculus to become a better modeller.

I hesitate also at having photoshop rework what shouldn't have happened in AM but the creases I really meant was the occasions when you can't get away from a crossover and dead-tail spline the ")(" you mentioned. I tend to try to work on as low a patch count as possible and there isn't the luxury of re-configuring to elegant solutions always.
I would argue that if I created a human model with say 3000 or more patches, I should ALWAYS have reasonably elegant splinage as there are enough patches I could blend difficult splines into.

I also tried porcelain and that often gives you strange effects on some 5 point patches.
My initial point was to look at any "within AM" solutions that could be found for the future - surface approximation or surface continuity tools that could cover or deal with creases. i'd certainly pay to see a slider feature on the porcelain mat.

daft
02 February 2003, 03:50 PM
I find it more than strange that Hash hasn´t been able to fix creasing in A:M.
One of the reason for I bought A:M in the first place was because I had seen and tried surface tools in Max. This was few years back when surface tools were the best thing in max for making organic shapes. Surface tools are still today much smoother then Hash patches. Though ST don´t give the option of making 5 point patch, you can still make 3 point patches that don´t crease as much as they do in A:M.

Since I started using A:M, I haven´t seen Hash making any effort to fix this creashing problem. This makes me very frustrated. If they can´t do it. why not give a better support for importing poly´s? Which could also be rigged and animated. Maybe then we could get renderings that don´t look like crap!
This would aslo open up a whole new world for the A:M community.

I have heard that Hash had T-shirts while back, which said something like " say no to poly´s". The Ironic of the whole thing is that Hash patches get still calulated in polygons at some time in the process. If not in modeling then surely at rendering times.

hoochoochoochoo
02 February 2003, 04:28 PM
[QUOTE]Originally posted by daft
The Ironic of the whole thing is that Hash patches get still calulated in polygons at some time in the process. If not in modeling then surely at rendering times.

QUOTE]

Thats true, all AM models are rendered as polys at the end. If poly based apps have found ways to smoothen surfaces or even areas of surfaces at the end why not in AM?:shrug:

ypoissant
02 February 2003, 05:44 PM
A:M subdivides the patches at render time until it is smooth enough. This is an adaptive subdivision technique that can subdivide way much more, at some places, than normal poly models or even the subdividing scheme of sub-D. "untill smooth enough" means that the subdivision process will keep going until the polygonal approximation are close enough to the actual patch surface.

Thus, if your patch topology is organized in such a way that it would produce creases, the subdivision process will keep subdividing untill the polygonal mesh conforms to the crease that is modeled by the patch.

The assumption is that the renderer will render whatever the artist have put in the model, creases included.

This is never the case with poly models because there where no patches to conform to in the first place. But a polygon modeler have to deal with the idiocyncrasies of polygon renderers too. After a while, a poly modeler don't even think about it but have learned to avoid certain types of mistakes. Those mistakes where recognized through the renderer feedbacks and solution where integrated through trial and error. With time and only with time does a poly modeler become excellent. Achieving the craftmanship of a Paul Steed with polygons for instance, takes time and experience (and talent of course).

As for crease producing modeling techniques in A:M such as crossover, dangling splines, dead-end splines, flaring splines, etc., in my experience, they are all avoidable in any models whatever the patch density either through spline rerouting or with hooks. I even believe there is a tutorial about exactly that somewhere on the Sherwood forest.

If you where to use NURBS (which BTW means "Non-uniform Rational B-Splines") tools, for instance, you would not have those particular problematic constructs because the paradigm doesn't even allow them to be constructed. You would have to deal with patch set discontinuity though. You would also have to tweak 16 CVs per patches.

The A:M way of modeling is much more flexible. This flexibility comes with a price though. The price is that there are much more ways to make mistakes in the mesh topology. But A:M is forgiving in that it allows the modeler to make those mistakes and even renders them (with creases though but at least it renders them). After all, is it the application role to decide if it is really a mistake of if it is really the well thought intention of the modeler?

Yves

daft
02 February 2003, 06:35 PM
Hey Yves!.

Your tutorials and plugins have helped me alot, your side is much so appreciative.

I have to disagree with you on the point that "A:M way of modeling is much more flexible". Don´t really take it as flexibility to be able to mess my
model up more with tweaking. In poly modeling you have the ability to move points around without messing up the mesh. In A:M on the ohter hand you will definitly get more problems by moving a point on the model, and then it all gets downhill from that one move you made. You will maybe get a nice looking model 60 hours later in A:M.
While you have finished your poly model in 2 hours.
In my experience subdiv modeling is much more flexible. If I don´t like how edge is flowing in the model, I can just delete it with ease and cut my model up in another way.

Try to make a hole on a curved surface in A:M! so it looks smooth:)

ypoissant
02 February 2003, 07:50 PM
draft,

Since this thread started comparing A:M patches to Bezier or NURBs, and the implication that some other types of curves where intrinsically better than Hash curves, I was trying to stay in this type of paradigm.

By flexible, I meant that the A:M modeling paradigm allows you to build topologies that would simply not be even attemptable in other curved patch paradigms.

Poly modeling is easier to grasp because it involves first order thinking so to speak. You build your surface vertex by vertex. If something looks funny in some area, you just tweak the vertex near that area until you have the appearance you are looking for. The polygonal paradigm is extremely simple. They are all flat faces. You just need to try to align the flat faces until you get a smooth surface.

Curved patch modeling involves some second order thinking so to speak. Building patches requires more thinking than polygons. Each patches appearance is controlled by the corner CPs but also by the tangents. That's the nature of the beast. And sometime patches seems to have a life of their own because changing one corner tangent will affect the whole patch curvature.

With poly modeling, there are no concept like edge crossover, or dead-end edges or flaring edges. Thus, those are some types of modeling errors you would make in A:M that have no equivalent concepts in the poly world.

I'm not trying to argue that A:M is the best tool all around. I know very much that there are some models that I would not atempt with A:M. For instance, if I had to model a squared beveled hole in a sphere, I would have better luck with other software although it still is possible to do it in A:M. What I'm saying is that A:M have its own modeling paradigm and it's not all bad nor all good but it's certainly not all bad. If you take the trouble to learn how A:M works than you can get what you want most of the time.

Sure, it you start moving around points on an A:M mesh without understanding the implications, then there is a good probability that you will get nowhere. If you understand the implications and work with that in mind, then you raise your chances to get the results you are looking for. One of the implication of the Hash paradigm (and any other form of curved surfaces BTW) is that moving points around is not enough (although A:M does a better job than other curved patch technologies on this matter by doing the right thing most of the time). You also have to move the tangents.

Sub-D is another story altogether. It is truely very flexible. It also involves only a first order thinking when modeling. Sub-D is not without problems though. This is a very young technology and it comes with it own set of idiocyncrasies. Research is still going on to find ways to remove some of its limitations. Ironically, one problem with Sub-D is the current inability to control normals on the mesh. Research is going on to add this ability. I guess, with time, Sub-D will get as complex as NURBs. :shrug:

Yves

iii-d
02 February 2003, 09:30 PM
hi all,
can anyone tell me the difference between Sub-D and NURBs?
both build a cage around the wanted geometry?
i only know splines ;-)

oliver

ragtag
02 February 2003, 09:51 PM
Here's a method to get rid of creases.

1) Create your model, texture it and so on.
2) Create three versions of your model.
A) Normal model with textures
B) Same model, but with porcelain/average normals applied
C) Same model, but with a new b&w texture applied and a constant material. The black areas are of the texture are those where you want to show model A, and the white where you want model B...greys will be a mix of the two.
3) Render the scene three times (sigh...what about my render times:bounce: ), once with each version.
4) Now in your favorite compositor or image editing software, you combine image A and B, using image C as a mask.

This gets you the details where you want them and a smooth surface where you don't want any creases.

Ragnar

p.s. Disclaimer: This is all theory, but I'm 99.9% sure that this works. :)

koon69
02 February 2003, 10:09 PM
Is there a way to work around creasing? Yves said something about tutorials - which ones? Hash is working on this - what exactly? Thanks.

ypoissant
02 February 2003, 11:27 PM
I was pretty sure there was a tutorial concerning no-no splineage habits. I whent to Sherwood but couldn't find what I had in mind. Sorry. I might concoct a quick one if there is a need.

Yves

Kevin Sanderson
02 February 2003, 11:48 PM
Originally posted by ypoissant
I was pretty sure there was a tutorial concerning no-no splineage habits. I whent to Sherwood but couldn't find what I had in mind. Sorry. I might concoct a quick one if there is a need.

Yves

There were two that dealt with creases but they are both dead links. :shrug:

koon69
02 February 2003, 05:14 AM
Now if only Yves can whip up a great tutorial showing us what not to do - the world would be a great place!

hoochoochoochoo
02 February 2003, 09:53 AM
Since this thread started comparing A:M patches to Bezier or NURBs, and the implication that some other types of curves where intrinsically better than Hash curves
Maybe we have a language misunderstanding Yves as that was NOT what this thread was about. Maybe I also pick up from your replies a sense of “a bad craftsman blames his tools” though I’m glad you recognise there are things that are problematic to model in AM. I could have reeled off modelling apps like FormZ and Rhino that handle certain things better.

May I plainly state, I like AM, it has allowed me to try things in a virtual world that I usually can only achieve in “real 3D”. There are occassions when I have problems though and I was asking if there might be other ways that AM could handle certain surfacing problems. I don’t need calculus, I’m a modeller – maybe I’m a crap modeller. Maybe it was other bad craftsmen whose requests introduced 5pt patches, hooks and porcelain. They certainly make a difference but there are still problem areas.
In history Artists and Craftsmen have often had to clag together or cobble up makeshift tools or implements to create their work when traditional tools didn’t quite give the required result. That I believe is why we have the variety of tools and equipment available to us today.

Please don’t think I’m having a particular “go” at you or Hash but I’m NOT impressed until your tutorial shows us all how to be great modellers.

ypoissant
02 February 2003, 03:52 PM
:surprised
I’m NOT impressed until your tutorial shows us all how to be great modellers.

This is either a very sarcastic response to my offer or putting a huge pressure on any attempt to write a tutorial. So I offer you a quick tutorial and two posts later, you tell me that you will not be impresed until it turns you into a great modellers? This is a complete turn-off. Sorry, I won't take that responsibility.

Yves

hoochoochoochoo
02 February 2003, 05:51 PM
I personally have no wish for war between us. Thank you for your reply. No such tutorial is possible.
My words on this thread at the beginning.

would Hash Splines still be as flexible if they had a "Bezier-spline" like ability and could that get rid of some (I know it won't get rid of all) of the creasing that sometimes happens?
just some thoughts - shoot me down if you wish.

Wasn't expecting shooting that seemed to be personal, I'll take as cultural or language misunderstanding. I'll end this now.:beer:

koon69
02 February 2003, 02:43 AM
Yves I dont think there was any malice intended. But a tutorial would be great - not trying to make any great modeleres - that is asking for to much - but maybe point some modelers in the right path!

daft
02 February 2003, 02:10 PM
I am hoping that someone can enlighten me on why 3 point patches render so badly in A:M. It´s one of the main problem I have in modeling.

And to those that are useing v10. Does Stitch and Maintain Curvature work well in practice. Can one add a point on a spline without changing the shape of it.
I find the example at hash to be bad, they should show some example on a curved surface. Like a sphere.

Maybe someone with v10 could post some examples!

Regards
Helgi

amsmf
02 February 2003, 03:39 PM
My understanding of why 3 point patches should be avoided in A:M modeling is that during the render A:M will figure out the smoothness of the mesh by triangulating the patches as much as needed in order to create a smooth appearance. What may be one patch on an A:M model may be seen by the renderer as dozens of little triangles.

A three point patch is ALREADY triangulated--it's one triangle, so A:M reads it a flat surface which of course will stand out as a problem on a face or other organic body. That's why its always better to figure out ways to avoid 3 point patches.

That's my understanding of it. If I'm wrong, someone please smack me.:rolleyes:

John Keates
02 February 2003, 06:10 PM
Stitch and maintain curvature do work well. Stitch saves a hell of a lot of time (for me), and although I havn't used it much, tests on maintain curvature with stitch work well too. The geometry stays in exactly the same place. However you have to keep in mind the fact that those bias handles have changed.

Hope this helps

CGTalk Moderation
01 January 2006, 11:00 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.

1