Modeling workflow for productions

after reading the post in the new user forum about nurbs vs subds I wanted to air this question...those were some really good posts ...and they seem to hit on things that I've been rolling around in my head....
what is approach that seems to work the best for you?
I've used CPS a decent amount.... first I usually start roughing out the basic shape and then add details with the poly tools. Is this a common workflow? What type do you really want to end up with nurbs, subds? Like said before, for productions polys is not always the end product because the resolution is not adequate. So what works best for the final type that will provide good
deformations without cracking around the joints and so forth and works best for animating. What are some basic workflows that you guys would recommend? Rough out with nurbs or polys and then use subds for the final texturing, detailing and animations? I know this question is a bit subjective but I'd appreciate any insight that you guys can give me.

thx!

This varies from studio to studio

If you are working with nurbs and you are getting cracking in Maya then you either did not assign a global stitch or lattice to keep the patches together or you need to turn on smoothedge or twaek the tesselations.

Anyway I feel like I'm justy repeating what I allready stated in the other post. If your model is a single mesh then you will not get cracking. Sub-d's and polys can be single meshes. If you are doing 3D painting then painting on a single mesh is easier than painting on a nurbs patch model. If you are going to work in games you are most likely never going to work with nurbs for final model output.

Personally I'm used to working with Nurbs, It is easy now to covnert tot he other formats from nurbs so I can go that way if I need to. I doubt you will find many studioes using CPS since it is an unsuppotted tool and most likely some palces have written their own solution. CPS is mainly for people who do not have access to Sub-D's and since Sub-D's are included in Maya 4.5 I imagine it's use will wane. It would be better for you to be familair with the native tools anyhow.

My advice is to learn all the techniques if possible and from there you will find which you prefer. It seems that you can pretty much use your preference and still output to the format you will need to for your work. If you were mainly working in games then learning the poly tools is very important. If you plan to work in film you should know how to work with nurbs and beable to deal with native sub-d's and probably beabel to use the poly tools to some advantage as well.

The reasons to use nurbs for starting your model is that you can trace images and use those profiles to create surfaces with. Personally I find them more creative to work with.. although if you are really familair with the other tools you can probably wield them as creatively.

Gmask w e realy need to get this discussion in the faq its starting to plunk its head up anew.

Yeah.. I keep having this sensation of Deja Vu everytime I think about modelling formats... that I just can't shake...

BTW where are you.. you allways do the night shift and I guess you are either a night owl or on the otherside of the planet.

No i do this daytime... Im in europe , Finland to be precise. So i guess im on the other side of the planet (EET or GMT +2 timezone). So theres a 10 hour shift between me and california for example.

Hi guys. Sorry to beat this dead horse but I've got a slightly different, and hopefully somewhat new, question to add to the issue.

I was doing some experimenting with patch modelling the other week, and it seemed to me to be nearly impossible to get a 'truly' seamless join between patches. I mean, I stitched two Nurbs patches with identical parameterization, checked in the wireframe to make sure CVs coincided, and used custom tesselation to ensure that the triangles making up the final rendered mesh lined up perfectly. Still, if I brought the camera close enough, there was always a hairline discontinuity visible. The shading was nice and smooth so the normals were lined up fine, but there was always that minute seam visible if zoomed in far enough.

So my question is: with NURBS modelling, is it possible to achieve TRUE seamlessness, or is 'a very very very good approximation' the best we can ever hope to get?

I must say I was surprised at how robust the seams were; I spent a LOT of time playing with stitching and tesselation settings to no avail. Don't get me wrong, they weren't visible from any 'normal' camera distance so they probably wouldn't affect 99% of projects... But I'm interested in the principle of the thing.

Thanks for any wisdom,

-Steve

I just spent the past several days make seam after seam on a very geometric shape with lots of mergeing angles and curved surfaces. I had no problem making it seamless but spent alot of time rebiuilding surfaces to get the shape that I wanted or to just figure out how the surfaces should meet. Keep in mind that the view you have in the GUI of the surface is not as accurate as it will be when rendered. If you isoparms are continous and meet at all edges then tesselation should not be an issue.

You can increase the display resolution for surfaces but it should not be necceasary. Also it is constant battle to fight how Maya decides wether or not to make a surface uniform and it makes mistakes.. I guess they are mistakes. Anytime you attach surfaces you have to rebuild it with CV's so that it doe snot deform in some weird way. You need to delete history often. And when all else fails and a surfaces does not act as expected you have to export it as dxf and reimport it. When you connect to surfaces and you do not wish for it to lkose the current curvature you need to connect without blend and then rebuil dwith the correct number of UV's in the direction of the attachment.

I have been using global stitch alot more to tighten up problem areas which great success you just need to use settings that will not cause major distortion.

hmm seems to me that you had the wrong tesselation settings on, you can achive very exact tsselation coencides if you so wish (if the tesselation casts with the last or second last floating point accuracy is of no consequence since no one will notice by the way, but this should be very accurate). First set the view render tesselation on, then turn explicit tesselation on, then disable the chord height ratio. Then set the uv values to what you wish. Even other combos are possible...
But i asure you wrong tesselation parameters will cause you problems. By the way when i tested this i actualy got perfect coincides within 0.00000001 units atleast.

Gmask and joojaa, thanks as usual for your expert advice. I just sat down and tried to reproduce the seams I was getting earlier and was unable to do so--I guess I was doing something wrong before, but I'm not entirely sure what!

I'll continue to explore (best way to learn, after all), and if I come across anything that gives a more concrete example of what I was talking about, I'll let you know.

Gmask, is there any chance I could get a look at that patch model you've been working on? --I'd be very interested in seeing exactly how you go about wrangling those seams. Of course, if you're not able or willing to just hand out your work like that, I understand completely.

Thanks again, and best regards.

-Steve

Steve... It's the model that I posted under Tire for a Classic Mercedes in the Pro section. I was going the wire frame but it's really dense .. too dense. For human modelling I allways refer to the models at www.zoorender.com as they are suberb.

I learned alot while modelling this wheel. If at all possible it is nice to plan your intersections just like you would for a face. Then upon determining that you can build your patchs so that they are orderly.. For my model if you want to punch a hole in something it is best if the hole is dead center between isoparms so that no isoparm will be further away from the edge of the hole than another. The process is called socking.. I don't know why though.

Anyway take a look at the Tutorials in the Maya Living FAQ.. I have added even more to the modelling section and most of what I know is there.

The important thing to remember is that working with nurbs is more about process than the tools themselves. WHne attaching and detaching surfaces you never can predict how they mught end up. They may end up chord length which is bad. The surface direction may be incompatible and no matter what you do the patches will not connect until you redirect one of the surfaces.

Sometimes the only way to correct a surface is to reselect the isoaprms wher ethey are supposed to be and reloft the patch.

I will try to put together a quicky tutorial on making corners with patch models and furtehr socking tricks as I think that will give alot of insight into the process.

Please don't go to any further trouble for my sake... you've already been more than generous, and my post was more of an idle wondering than a pressing issue.

It's been a while since I looked at the Living FAQ, definitely a lot of new stuff there for me to pore over. I'll let you know how it goes.

Thanks again,

-Steve

Here's a modelling tutorial.. It's not great but has some good tips,..

http://www.gmask.com/tutorials/corners/corners.html

the visibility of seams (not tesselation cracks, or stitch samples) - little shadding ripples that break perfect surface continuity - is a problem that arises when you are trying to achieve continuity across regions of high curvature. if you align with curvature continuity, you'll find lots of these will dissappear (along with a surface flowing in a manner you probably dont care for). you'll find these seams are more evident if you apply a specular shader, a pleasant surprise after having modelled with lambert. the (only?) way to comabat this is to arrange your patches as best as possible to place seams in areas of low curvature and deformaiton, or hidden. + texturing to break up the seam visibility. its a limitation, like nurbs topology - deal w/it. this is the way i see it. love to hear more thoughts on this.

(I've been unable to access HighEnd for the last couple of days. Has the site been down, or is it just a problem with my connection?)

Gmask, I just read over your tutorial. Great stuff, and a darn sight better than the built-in 'Round' tool, that's for sure!!

Campbell, good notes to keep in mind. Seam shading problems are NOT the same as tesselation or sampling problems, and of course render tesselation and viewport tesselation are different kettles of fish as well.

As for more thoughts... what I've been wondering about is how flexible patches are in terms of what patterns of multiple patches can be joined together.

I downloaded the Adam model from Zoorender (gmask was right, it's superb), so let me use some screenshots of it as illustrations.

For example, it's straightforward to join four square patches at their corners, so that they make a large square with a patch for each quadrant, like so:

But what else is possible? If you make the isoparms line up, there's no reason why you can't stick more than four patches together at their corners, like this:

--But I was surprised, when examining this Adam model, to see that there were seamless joins of multiple patches in the middle of the edge of one patch, like this:

Obviously 'Stitch Partial Edges' is an option useful in such a case, and the intermediate curves involved need to be specified numerically for best precision. And this needn't be done in a neat 'T'; it can also be made into a crazy angled junction involving four or more patches:

So obviously there's more to joining patches than just lining up corners and edges. Which is good, because the topology of the human face, among many other things, means that NURBS would do a poor job if they always had to be joined four-together. So where are the limitations? Does anyone know if it's possible to join three patches at their corners, and end up with a seamless surface? --I tried it and couldn't quite manage it, but I'm no expert. What about joining multiple patches, two or more along partial edges, with others meeting at their corners, or... or... or...?!?!

This technique obviously goes quite deep. I for one am going to have to spend a lot more time exploring it... thanks again to everyone who contributed to the thread or the FAQ to try to help the less experienced among us along.

-Steve

The website has been having some problems lately.. I don't really know why :frowning:

Gmask, I just read over your tutorial. Great stuff, and a darn sight better than the built-in 'Round' tool, that's for sure!!

Thanks.. I never use the round tool.. I didn't even think about once while I was writing the tutorial?!?!?!


Well as long as the isoparms line up you can either do the partial egdge stitch or in this case it was probably a global stitch.

You always want to have 4 if possible and 5 is recommended over 3. It is not advised to have more than 5 patches meet. Although in the ca eof the forehead ther ecould have been 6 patches there but he decided to merge the two nose patches after mirroring the body. You could probably merge most of the patches falling on the symmetry line.

You generally want intersections like this to meet in a flat place.