A couple of questions about hard-surface best practices...

[El Burritoh]

I have a couple of question I'd like ask our hard surface modelers. I know John has been posting around here so maybe he has some insight... Anyone else who has experience to share, please do so!

I'm a modeler at a studio in the U.S. and although I "can" model just about anything, I do better with hard-surface assets.


1) Edge Weighting

Whenever I'm sharpening up hard edges, I always use actual geometry to do so, by either adding edge loops or sections of edge loops. I've been tempted to use edge weighting, but it never works into my workflow (where geometry will pass between apps and the edge-weighting data would be lost). In the context of a larger studio like Weta, how are edge-weights viewed? Are they frowned up for pipeline reasons (or other?), or are they a perfectly acceptable modeling tool?



2) N-gons

This is still related to hard-surface modeling... I have always heard that n-gons are of the devil and must be avoided at all cost (and for organic stuff, sure). But in actual practice, for surfaces that aren't going to deform, I have found that some 5-sided SubDs actually render better than if the patch had a triangle in there somewhere. It's not always the case, but frequently I will try to make a section all quads and tris, only to delete an edge and find that as an n-gon, the tension balances out and the darn thing actually looks better! I have heard from other hard-surface modelers in the industry that n-gons on non-deforming surfaces, especially if the surface is flat, is okay in many situations. Is this the sort of thing that depends on the studio? Or now that Catmull-Clark type SubDs are the norm, are n-gons pretty much okay?

[collings]

here is a cool video from Greg Petchkovski about topology to avoid smoothing artifact etc.

http://vimeo.com/2158706

[El Burritoh]

Thanks, Nicolas. I've seen that vid before and it's very helpful, especially for the puzzle work of getting rid of unwanted 3-5-sided polys. Very good stuff.

I guess I'm just wondering if the use of n-gons is strictly prohibited in some studios, or if it depends on the usage of the model, or if there aren't any artifacts on a hard-surface model, does it even matter? I know the discussion is still ongoing as technology progresses. Just wondering what the philosophy at the higher levels of the industry is.

And I am referring specifically to hard-surface modeling because those surfaces don't really deform like organic models who need to be rigged to flex and bend.

[El Burritoh]

Anyone else?

[JohnSG]

Hey El Burritoh

1) Edge Weighting
I know of it but I rarely ever use it for Hard-surface models, And this is due to the reasons you pointed out, to many different part of a studios pipeline and to many different tools. Sometimes the simplest option is the best option. So I tend to lean toward keeping everything a geo.

So really it just down to the studios pipeline and the best way you work with it

2) N-gons
You do pick the dangerous topics . From what i have found and heard its very much down to either the studio having a preference or the modeller themselves.

Personally I try to avoid n-gons and just replace them with tri's where ever we can. There are some render related issues that can arise from using these as well.

So yet again its really just down to the studios pipeline or the modeller.
I kinda feel like I haven't answered your questions, but at the same time its very hard to give an answer considering the amount of variables, what use, what pipeline etc...
Just my 2c, Hope it Helps,
John

[El Burritoh]

Thanks John, that answers quite a bit actually!


You've piqued my interest about n-gons causing memory problems for spherical harmonics. I am familiar with spherical harmonics as a solution for image-based lighting, but I'm not familiar with how n-gons affect that. Is this an obstacle unique to certain rendering engines, or is this an issue with spherical harmonics in general?

[JohnSG]

To be honest I am not sure on the technical aspect of it.

[Billabong]

Hey thanks John, that did answer a lot, I would like to know why you guys sub-d your stuff at render times? is it because of Stand alone render engine?

-B

[JohnSG]

Its down to a few reasons, not all related to the Models dept. Cant speak for the other dept, But for us anyway, modelling hard-surface in polies is alot easier and quicker then true subd's. No reason to model in subd when your render engine will just convert polies for you? :)

[Billabong]

yeah that does make sense. I just didn't understand why you guys would subd it at render times.

Thanks so much

-B

[El Burritoh]

Subdividing at render time is handy for several reasons. Mostly, it speeds up scene performance. Many times in XSI, we'll have our objects display in straight polys but render in SubD. During the modeling process, we'll display things in SubD to check how it's working, but once it's passed on to the animators, unless it's a very lightweight scene/asset, we'll leave in polys and have it subdivide at render time. It's an extra degree of control over scene optimization, and with dense scenes you need everything you can get.

[ianucci]

What are the advantages of 'true' subdivs over the smoothed poly workflow? I've don't really know what true subdivs are and how they differ from polys to be honest. :o

[El Burritoh]

It's a little like bitmap images versus vector files.

Polygons are the bitmap version. You can do very well with polys as long as you have enough to define the shape you want. Like pixels, the more you have, the more detail you get. However, just like bitmapped images, you can only get so close before the shape starts to deteriorate. In the case of bitmapped images, you get pixelation. In the case of polys, you can get faceting. And smoothing angles don't solve the problem, they just smooth over things from a head-on perspective. Along the edge of the polygonal structure, faceting will eventually occur.

Subdivision surfaces are a little more like vector files (.eps, .ai). These can define a curvature at any resolution, and can adaptively change their subdivisions to smooth out polygonal shapes. It doesn't add geometry that the user can immediately interact with, but it does add geometry visually, giving a smoother curve to edges and all but removing faceting (almost). It allows the 3D camera to get much closer to an object before it starts losing detail.

Also, some shading effects don't work nearly as well on straight polygonal shapes, no matter what their smoothing angle is set to, leaving SubDs or Nurbs as the solutions.

That's the gist of it, at least.

[ianucci]

Ah right! Thanks for that El Burritoh, I didn't realise they worked in a similar fashion to NURBS, now I understand the appeal.

[mynameisnick24]

Sorry to be late to the party with the post, but is it always a good practice to convert organic poly-models to Sub-Div before render, or does it depend on the situation. If I'm understanding correctly it is, but I'm just making sure.

[JohnSG]

It really does depend on the situation,

All the hard-surface work I do is left as polygons, as there is far to much work involved in converting them to another format once you have made them in polies. Also each format polies, nurbs, subds, all have different effects on uvs depending on your render process / engine and how they are converted. So basically leaving it till render time to convert the geo is more productive and efficient.

[mynameisnick24]

Cool, Thanks for the Info John.

[El Burritoh]

That's very interesting. I guess your models are pretty dense then? Take the military mech-suit from Avatar, for example. Was that all polygons, no subDs? What was the polycount on it?

[JohnSG]

Yes my models can get pretty dense. Personally i think keeping everything as poly's helps to reduce the amount of time involved in creating each assets.
It kinda sucks that I cant post any pic's or wireframes of the Amp-suit due to our NDA's. But you know how it is.

[Billabong]

pretty high. Doesnt make me feel to bad about having such hi counts any more.

Thanks John
-B

[El Burritoh]

That's good stuff to know, John. Thanks! It's good to have the perspective of an inside man, so to speak.

Doing everything in raw polies definitely frees one from those peculiar mental gymnastics unique to SubDs and lets you cut right the chase.


I really wish you could post some wires of the Amp-suit. It's unfortunate that even now that the film has been released, your NDA still prevents that. Ah well.

[JohnSG]

Quote:
Originally Posted by Billabong
wow. 2.5 mil huh? pretty high. Doesnt make me feel to bad about having such hi counts any more.
Thanks John
-B

eh don't go to crazy now buddy

The weight doesn't matter so much as long as it is where it is needed. So its detail where you cant avoid it.

[JohnSG]

That's good stuff to know, John. Thanks! It's good to have the perspective of an inside man, so to speak.

Doing everything in raw polies definitely frees one from those peculiar mental gymnastics unique to SubDs and lets you cut right the chase.

So John, the trouble you mentioned previously about n-gons and spherical harmonics...those were pertaining to all-poly meshes then?


I really wish you could post some wires of the Amp-suit. It's unfortunate that even now that the film has been released, your NDA still prevents that. Ah well.


Yeah SubDs are not the greatest thing to try and figure out, as easy to explaining as female logic. :D

Yes the previous ngon question was relating to an all poly mesh.

Yeah NDA's suck big time. :rant:

[Billabong]

eh don't go to crazy now buddy :thumbsup:
.


HA! its really hard not to though,(j/k) I get what your saying though.

Thanks

-B

[El Burritoh]

John, for things like nuts and bolts, could you not use instances instead of actual geometry?

[JohnSG]

John, for things like nuts and bolts, could you not use instances instead of actual geometry?

Personally I don't use instances,

It also depends on where they are and how close the camera is to them, you may need variations in your bolts to take the cg look off them, which sometimes it just better off done in the geo then the texture. If its a super close up shot.

Not saying don't use them or anything, just that I dont.

[ianucci]

Hey John like the other guys I really appreciate your posts about this stuff. I hope you don't mind if I fire a few more questions at you :p.

I'm curious how many different level of detail versions you guys made of the Amp suit for example, I'm guessing at least 3 or 4?

When you do the different versions do you begin with the super high poly then tweak that to get the other versions?

You mention that sometimes you might do variants of small details like nuts and bolts for close ups, do you just make minor alterations to the overall form or do you have to go even further than that like maybe a pass in a sculpting app?

[JohnSG]

Yeah no problem, It is CG Feedback, questions are why we are here.

I'm curious how many different level of detail versions you guys made of the Amp suit for example, I'm guessing at least 3 or 4?
Its kinda hard to answer this one, Its very much depends on the project and level of detail in it. For a super heavy cg film the ideal meathod would be to created levels of details for almost everything. For instance 3 levels, Full res. medium, and lo.

When you do the different versions do you begin with the super high poly then tweak that to get the other versions?
Yes, generally this is the best approach, and I also tend to leave it until the hero is approved (first director approval anyway), although as some of you out there will already know, northing is every approved until its in the cinema

You mention that sometimes you might do variants of small details like nuts and bolts for close ups, do you just make minor alterations to the overall form or do you have to go even further than that like maybe a pass in a sculpting app?
For nuts and bolts, i would just tweak the geo a bit, going to a sculpting app is a bit over kill. Generally for hardsurface you can get most things without any sculpting app.

[El Burritoh]

I'm with ianucci: thank you, John for all this insight. I know it's how Weta does things, but it's always helpful to know how people are solving common problems elsewhere.


And now to continue with the grueling interrogation of John Galvin...

For something like the Amp suit, which is a front-and-center kind of asset, and where functionality and design are so directly linked, is the artist responsible for figuring out any of the mechanical stuff or is it (more or less) fully designed by the concept artists? In my own experience, I sometimes run into things that make me go "uh...yeah..that design's not gonna allow the right kind of rotation..." And it's up to me to fix it. But with something as high-profile as the Amp suit, is there a good back-and-forth dynamic between the departments when that happens, or is the Models department stuck with what it gets?

[JohnSG]

Quote:
Originally Posted by El Burritoh
For something like the Amp suit, which is a front-and-center kind of asset, and where functionality and design are so directly linked, is the artist responsible for figuring out any of the mechanical stuff or is it (more or less) fully designed by the concept artists? In my own experience, I sometimes run into things that make me go "uh...yeah..that design's not gonna allow the right kind of rotation..." And it's up to me to fix it. But with something as high-profile as the Amp suit, is there a good back-and-forth dynamic between the departments when that happens, or is the Models department stuck with what it gets?


Hey El Burritoh
I totally know what you mean. it is very much down to the Concept artist or designer. These people have a great eye for what looks cool, but sometimes some of them are not very technically / mechanical minded. So its kinda 50/50.

Sometimes I have received concept art that looks great in 2d, but cant function in 3d. This is one reason I tell people to try and do a few models from concept artist artwork, as its the very similar to working in film. Also its nice to see that someone can match a drawing either there own or someone else. All good things to see in a reel. ( that ones free) the rest will cost you
As for a High profile asset there you generally receive pretty detailed designs, But they are usually not fully tested in 3d. So the first step would be to built a test model very low, basic shapes and put it thought an extensive range of motion tests, If any problem arise there are two options, go back to the concept artist or its left up to me. The majority of the hard-surface team here are all very technically / mechanical minded, and all love the challenge in creating and designing solutions for these sorts of problems.
So to answer your question, Yes it can be quite similar to what you are used to.

[Billabong]

I just have to say, this is a fantastic thread. Learning a lot here

-B

[musashidan]

I just have to say, this is a fantastic thread. Learning a lot here

-B

I'd say it's more akin to a John Galvin Q&A session at this stage. :)

And in keeping with that.............I wondered, John, would you use a similar all poly/non sub-D workflow on personal projects? And although the NDA on work stuff applies maybe you could post a few wires of personal project work that might have similar wireframes to the Amp-suit.

Cheers.

[JohnSG]

Hey musashidan

Nice to see someone from my home-town. :beerchug:

To be perfectly honest, it has been quite a few years since I have done any personal work at all, I might have one or two car that are half finished but thats about it. Avatar was pretty brutal on us, Alot of hours for a very long time. And any personal work i did before that I would be to ashamed to put up now. Sorry :(

I have no problem in critiquing work-flow or wire-frame mesh's that have been posted, and if there are any issues or things I could see to alter or improve, I would have no problem in demonstrating some techniques and posting parts of meshs. :thumbsup:

[musashidan]

Hey musashidan

Nice to see someone from my home-town. :beerchug:

Even nicer to see lads from our hometown working for WETA. Good man for flying the 'oul flag. :beerchug: Amazing work on the movie.

I have no problem in critiquing work-flow or wire-frame mesh's that have been posted, and if there are any issues or things I could see to alter or improve, I would have no problem in demonstrating some techniques and posting parts of meshs. :thumbsup:

I was just intrigued that the inorganic modeling was all poly/smoothing angles and no Sub-D whatsoever. It would be really terrific to see some screengrabs of you demonstrating techniques/posting parts of meshes.
I'm sure it would be well received by alot of the lads here.

cheers mate. :thumbsup:

[JohnSG]

I was just intrigued that the inorganic modeling was all poly/smoothing angles and no Sub-D whatsoever. It would be really terrific to see some screengrabs of you demonstrating techniques/posting parts of meshes.
I'm sure it would be well received by alot of the lads here.

cheers mate. :thumbsup:

I will see what i can do. I am a bit flat out at the moment, alot going on.

[musashidan]

I will see what i can do. I am a bit flat out at the moment, alot going on.

I'm sure you do. Cheers man.

[tiktok]

More than ngons or subds, I'm more curious about bevels. Everytime I do hardsurface things 90% of my time is spent going back through and beveling everything. I've heard people say never to use single bevels, to always edge loop both sides of the hard edge, though often I find many uses for single bevels as well as the like. Also, what about use of bevel shaders? Is that commonly used, or not?

So, what does everyone do about bevels here? and what techniques do you use to get your final result?
:sw:
-Bayard

edit: I suppose what I'm im referring to would be subd bevels.

[JohnSG]

More than ngons or subds, I'm more curious about bevels. Everytime I do hardsurface things 90% of my time is spent going back through and beveling everything. I've heard people say never to use single bevels, to always edge loop both sides of the hard edge, though often I find many uses for single bevels as well as the like. Also, what about use of bevel shaders? Is that commonly used, or not?

So, what does everyone do about bevels here? and what techniques do you use to get your final result?

-Bayard

I very rarely if ever use the edge crease for bevels in the same way you can on subds, It just takes to much time to set-up and tweak. And i have never even heard about the bevel shader.

The solution that I think gives the best result is triple edge loops, for a few reason, they are quick to make, easy to edit , and you can also add in varaination down the line to indicate damage or warn sections, This helps to break up highlights. :D

[El Burritoh]

Bevel Shaders aren't that useful in the end because 1) it restricts your pipeline and 2) the effect is too uniform and takes control away from the artist over how sharp certain edges really are.

Whether in polies or SubD, I always wind up edge-looping to control the edge sharpness, though sometimes I also use an "edge bevel" tool that adds any number of subdivisions to the curvature of that edge. Just depends.

[tiktok]

john: a pic of your 'triple edgeloop' thing would be great to see. Bevel shaders just mess with the normals and display the edges as if they had bevels on them.

el: Curious what everyone uses for plugins and such when hard surface modeling? nex, byrons polytools...etc? (or max's polyboost?)

[JohnSG]

john: a pic of your 'triple edgeloop' thing would be great to see. Bevel shaders just mess with the normals and display the edges as if they had bevels on them

Yeah no problem, I will post some early next week as I am away for the next couple of days. But as soon as i get back i will try sort out some examples

Curious what everyone uses for plugins and such when hard surface modeling? nex, byrons polytools...etc? (or max's polyboost?)

As for plug-ins, sounds like you are using max, Been a long time since i used it and i don't remember the poly tools being great then, not up to date with the latest versions.

I mainly use maya and for most hard-surface poly modeling I find i don't need any more then the basic poly tools they have.

[El Burritoh]

I recently started using Max for FumeFX at work, but beyond that I don't use it much. But I would like to know what all the new graphite tools offer!

Not to make this a software debate, but for straight-up modeling I use Modo. It really is like a 3D swiss-army knife. Our studio is primarily an XSI house and although XSI has good modeling tools, modo is really hard to beat in that area and I find it much more fluid. Although I've also started learning Maya of late and I could see myself using Maya just as much. Maya and Modo share a surprising number of methodologies.

I think I know the technique John is talking about, but I wouldn't turn down screenhots.

[mynameisnick24]

Could someone let me know what a bevel shader is. Is it exactly as it sounds, A shader that fakes a beveled edge on a piece of geometry, or is it something different.

Also
Curious what everyone uses for plugins and such when hard surface modeling? nex, byrons polytools...etc? (or max's polyboost?)

When I worked in Max I used the Polyboost plugin for modeling. It had some very nice tools to speed up workflow and it was relatively cheap. But it tended to be little buggy at times, this was over a year ago though so it may have been improved.

[El Burritoh]

I've personally never used one but I've seen them floating around. Depending on the rendering engine (I'm think the ones I've seen are for MR), you can get shaders that will do pretty much what tiktok said: render hard edges in such a way that they catch light and appear rounded/beveled/chamfered, although the geo doesn't change at all. XSI and Max have some, and I'm sure there's something for Maya too. But it's just not a proper way to deal with edges (for the vast majority of tasks, because I guess the word "proper" is relative).

My understanding is that the Polyboost plugins were fully integrated into Max 2010 and got renamed as the Graphite modeling tools.

[ianucci]

I've used the graphite/polyboost tools a little and there is a tremendous amount of power within them, but I think the UI for them could be streamlined somewhat. I can still model a lot quicker in Cinema despite its aged modelling tool-set, but then I am very much a newcomer to max.

I haven't yet tried Modo, must give it a look. I've used Silo, and found that a lot nicer to use than Max and Maya.

[mynameisnick24]

Thanks Mr. Burritoh for the explanation, I assumed that's what it was but just wasn't sure. And also for updating me on max it's been a while since I've used it.

Ian, I agree with you in terms of the UI, It was organized very weird some, tools were buried in other tools which definitely was confusing at times. At least it was like that in the Polyboost stage. I'd imagine it's still somewhat the same.

[JohnSG]

So I managed to find a WIP car, Its still not finished, quite a bit still to go, but it will give an good indication of the poly mesh.

721

Any questions please let me know, or if there are any areas of the car you would like a close up off let me know too

[Billabong]

very nice John, very nice.

I meant to ask you this a while back and I cant remember if its been asked yet or not.

I know you said you guys work in polies. but since Maya's Nurbs are really great(at least thats what Ive heard) do you ever start out with Nurbs and then convert it to polies?

If so, what would be the circumstances of using Nurbs

-B

[tiktok]

thanks much for posting this John! Would be great to see close ups of any of the complicated areas, vents, valve covers...etc. Anything to see bevels as well, and how you dealt with them.

:bouncy:
-Bayard

[JohnSG]

I know you said you guys work in polies. but since Maya's Nurbs are really great(at least thats what Ive heard) do you ever start out with Nurbs and then convert it to polies?
-B


Yes I do sometimes start in nurbs and convert to poly's. It very much depend on the object you have to make, and how complex it is. The good thing with nurbs , after you convert them to polies you are left with a pretty nice flow across your mesh. Good for planes and large curved surfaces.

For cars i tend to just start from polies, super low shapes and built them up, as the nurbs surface for some of the body panel shapes are too complex to build in nurbs and convert as they leave you with a weird surface which then needs to be edited anyway. So i find starting in polies quicker and as you have more control at the start it help to keep the mesh resolution down. :thumbsup:

[JohnSG]

thanks much for posting this John! Would be great to see close ups of any of the complicated areas, vents, valve covers...etc. Anything to see bevels as well, and how you dealt with them.

:bouncy:
-Bayard

No Problem, I will put together a list of detailed areas and try to post them later today.

[Billabong]

Yeah I do think that's one of the great thing about Maya, is its Nurb capabilities. Max's are horrible and basically unusable.

Thanks John

-B

[PascalR]

No Problem, I will put together a list of detailed areas and try to post them later today.

And I'll probably plug this thread on the front page tonight :) Great model John, thanks for sharing your tips and tricks here! :thumbsup:

P.

[El Burritoh]

Thanks for posting this John. I recently did something similar as your nurbs method, but I had modeled a rough of the object first in polies. Then to make sure I had a good even shape I used the polyonal model as a constraint and modeled in splines on top of it, then patched those. I'm not much of a nurbs modeler, and that's something I intend to correct. Using a polygonal structure as a foundation and then modeling with nurbs/splines on top of that adds extra modeling steps, but gives much better results!

As for the mesh you posted, I wanted to ask you how you solve the tension variation in situations like the one below. The corner is well-defined, but the edge between it and the adjacent corner below it smooths out a bit much. Would you just add some edge loops in there to tighten up that middle section of the edge?

http://timcrowson.smugmug.com/photos/794716754_M5KA3-O.jpg

[JohnSG]

As for the mesh you posted, I wanted to ask you how you solve the tension variation in situations like the one below. The corner is well-defined, but the edge between it and the adjacent corner below it smooths out a bit much. Would you just add some edge loops in there to tighten up that middle section of the edge?
[/IMG]

As i mentioned it was WIP, so there are still areas that need to be tweaked. I did have to add one more edge loop just around the front to tidy up the piece and stop the stretching.

722

There would be a few thoughts on this section, I wanted to get the basic shape with the least amount of points, then after the first test render if its an issue they i would treat the area with a bit more geo.

Alot of time can be wasted over thinking things that are not even seen, so for small details like that i always try a low ammount of geo first.

[tonytrout]

Im stunned at how few polys you get away with, your model was a bit of an eye opener. Thanks for sharing your methods, much appreciated.:thumbsup:

[JohnSG]

Not a problem, Glad to help out. :thumbsup:

Just encase anyone else is wondering the current poly count is 40k.

[JohnSG]

Here are the list of image I promised.

Something to note, some of these techniques may seem unfamiliar, and may not work with your pipeline. Feel free to ask any questions.

740

[tiktok]

John:
This is so great to see!. Really suprised how rough everything looks yet how clean it is subdivided! :bouncy:
very informative!!!
-Bayard

[ianucci]

Nice stuff John. I have a couple of questions.

That last shot, i see you have some edge loops around some of the segments so presumably that part is not meant to smooth perfect round. Be cool if you could show some close ups of the smoothed surface.

How do you decide when you can get away with using intersecting geometry, that has always been a tough one of me. I presume on small parts you can get away with it because they wont be seen close up, does that mean beyond a certain size everything has to have smoothed edges? I'm always concerned about that 'infinitely sharp edge' with hard surface modelling.

[JohnSG]

Yes you can get away with quite alot. Best way to think about it is that you are modelling with Bézier curves. The minimum amount of geo possible to create the required arc or shape.

[El Burritoh]

I've always have problems with "stars" (high-valence points) on curved surfaces. Is that a problem for anyone else?

[JohnSG]

Here is a few images showing a few tips in how to create really clean light geo.

As i mentioned the best way to think about it is Bézier curves, and the curves maya already has are great for helping to layout your geo.

Take this example, you want to cut a hole into a surface, Take a curve and using the cv points from a default circle (8 is a good number 6 and 4 can sometime be to low and stretch) using the split poly tool I cut in my points roughly and then snap them to the circles cv point, now you know that when it sub divides it will be a perfect circle.

We use this technique alot here, from circles to very complex shapes, I regularly draw the outline of the shape I want in curves to see how many point to need to create it.

741

[JohnSG]

I've always have problems with "stars" (high-valence points) on curved surfaces. Is that a problem for anyone else?

It can be down to each studios pipeline, What could be an issue for one place may not be to another. Do you have an image that would show the problem you have with them.

[rasmusW]

wow! this is exactly what i've been seeking around other forums. people, ranking high and low shareing workflows. you don't see many forums doing this imo.
i'm learning so much from this. thanks john and the rest of you guys.

here is a few of my thoughts on some of the topics covered here (in no particulare order):

at work we have used the mr bevel shader (roundcorner). though i wouldn't normally put my faith in the shader to do the work. it did work well on that particulare lego project. we did it due to time constraints and art direction.

as for apps i use for modeling. at work it's all maya and zbrush, but at home i mostly use silo. it's very basic toolset, but it just works so well. especially for organic modeling. for hardsurface modeling i'd probally lean toward maya.

-r

[JohnSG]

Nice stuff John. I have a couple of questions.

That last shot, i see you have some edge loops around some of the segments so presumably that part is not meant to smooth perfect round. Be cool if you could show some close ups of the smoothed surface.

How do you decide when you can get away with using intersecting geometry, that has always been a tough one of me. I presume on small parts you can get away with it because they wont be seen close up, does that mean beyond a certain size everything has to have smoothed edges? I'm always concerned about that 'infinitely sharp edge' with hard surface modelling.

Its a fairly small piece on the bonnet. I try not to over do it in areas that will not be seen unless it was an extreme close-up. If this was a bigger piece or seen more closely I would have increased the amount of geo around the outside, which would result in a smooth surface.

[El Burritoh]

There's also another way of putting round cylinders into meshes like that. Either as cylinders that come off, or holes that go in. You can grab those four initial polies, inset them (bevel without shifting out), and collapse them in on a single point. Then bevel that central point out and you've got a circular shape to do whatever you want with. Add edge loops as needed to tighten things up.

I don't have any examples of the high-valence point problem that I can post, the situation you have around the interior ring of your wheel rims is an area that would probably give me rendering errors. Depends on the rendering engine maybe?

[weiluntsai]

http://www.youtube.com/watch?v=kIfvbKovHWY
very smart way :bouncy:

[El Burritoh]

Arrays like that are annoying. There's no quick way to do that stuff. The snapping in that app (what app was that?) seemed very responsive!

Reminds me of a chair (http://www.dynamiclens.com/files/ironchair.jpg) I did for a music video (wireframe of the seat (http://timcrowson.smugmug.com/photos/795564617_YcuSj-O.jpg)). And then it was used as a BG object! Argh!

[weiluntsai]

Arrays like that are annoying. There's no quick way to do that stuff. The snapping in that app (what app was that?) seemed very responsive!

Reminds me of a chair (http://www.dynamiclens.com/files/ironchair.jpg) I did for a music video (wireframe of the seat (http://timcrowson.smugmug.com/photos/795564617_YcuSj-O.jpg)). And then it was used as a BG object! Argh!

hummm...i thought you post a picture. nice job of the chair :beerchug:

[Billabong]

Arrays like that are annoying. There's no quick way to do that stuff. The snapping in that app (what app was that?) seemed very responsive!


Its Maya


The guy who made the video won Dominance war 4 on game artistians. He shows how he made the entire character. Watching these videos made me almost give up on 3D( just kidding), but seriously this guy has got some talent.

here are both parts of his making of. If you download them, they are higher quality

http://vimeo.com/5738519

http://vimeo.com/5734031

-B

[dr.kim]

Wow~ What a useful thread. I am modeling hard surface model right now and this thread couldn't be more helpful. I have question about the "stars" like El burritoh pointed out. I find that when that section gets smoothed, it'll make weird deformation to the UVs. Especially when there isn't enough GEO. Anyone else run into this problem?

[collings]

i had seen that making of some month ago, mind blowing stuff, very informative !
This guy is a killer clap

[El Burritoh]

Billabong, that guy is amazing. He pulled out the stops on that one. Stuff like that makes me feel like a noob and motivates me to keep going!

[Billabong]

Billabong, that guy is amazing. He pulled out the stops on that one. Stuff like that makes me feel like a noob and motivates me to keep going!

Yeah, I almost threw up when i saw this for the first time.

One day Tim, one day

-B

[PascalR]

very impressive work indeed clap

[El Burritoh]

It's interesting how he uses Zbrush to extract geometry, keep it low-res, and then bring it back into Maya. That's a smart way to avoid some complex modeling.

[JohnSG]

Thats nice work alright. Good attention to detail. clap

[Manuel Poehlau]

wow .. great videos. Hats off :thumbsup:

[clay]

Wow, that video was pretty incredible. Love the attention to detail and the working mechanics of it all! That's some modelling at its finest :love:

[musashidan]

Wow this thread has really taken off. Some really excellent stuff on here. Thanks John for posting up those wires. Some very interesting methods indeed.

[madmatt]

Very interesting topic and thanx for the links...

Bye

[iskon]

Absolutely agree with all you guys. That "DW 4 winner" guy is soooooo good at what he is doing. Plus he made making of for us. clap

[tingles]

Hey guys!

Saw this thread and felt i should contribute!
I recommend these video tuts by Grant Warwick.
Explains in great detail how to create good looking topology using various techniques, creating a variety of shapes.

http://digitalapprentice.net/Community/index.php?option=com_content&task=view&id=67&Itemid=154

Awesome thread, I'm a hard edge nut myself, so threads like these are a great read!

Marc

[collings]

tingles, that website is dead since a few month now. here are the working link for those videos, it's hosted on the autosdesk website the Area. you need to be a register user to access the tutorials.


http://area.autodesk.com/tutorials/inorganic_modeling_fundamentals_part_1
http://area.autodesk.com/tutorials/inorganic_modeling_fundamentals_part_2
http://area.autodesk.com/tutorials/inorganic_modeling_fundamentals_part_3

[tingles]

Hey thanks collings, my bad!

[ianucci]

I've recently been working on this Judge Dredd Helm. Just wondering if the topology looks ok and if anyone can suggest a way to remove the pinching that I've indicated.

1061

[JohnSG]

I've recently been working on this Judge Dredd Helm. Just wondering if the topology looks ok and if anyone can suggest a way to remove the pinching that I've indicated.

1061

The easiest way I think to treat this instead of adding way more geo to maintain the curved surface, would be to break this section out of the helmet and make it a separate slightly raised piece, This will able you to add more geo in a smaller area where need and give a little more depth to the model.

This pieces being the red band that runs around the helmet in the pic below

1062

[tingles]

EDIT: Had a sketch posted, but i totally screwed it ... would also recommend splitting the mesh :)

[ianucci]

Thanks guys, I'm always making things hard for myself trying to make things single meshes :).

[El Burritoh]

Is that modo? Looks like modo, smells like modo, tastes like modo...

I've encountered stuff like this and the most elegant solution I've found is to subdivide the entire mesh. Separating that area is a design decision, but also reduces the final number of subdivided polygons. I don't know of an elegant way to clean that corner up without increasing the total resolution of that mesh. So yeah, what they said...

[ianucci]

Its actually cinema 4D :). I've been meaning to have a go with Modo, Cinemas modelling tools have a few gaps that need plugging although it still has some nice tools I've not seen in any other apps.

El Burrito, If you were to subdivide and clean up that area, how would you approach that? Just tweaking the points with the move tool/brush/soft selection? I may try both approaches just to see which look I prefer because as you say it is a design choice.

[JohnSG]

I've been meaning to have a go with Modo,


Just my 2c, I have tired both of these packages and there are many features that I like about both and some I wish maya had. But the one thing i have noticed over the years is that alot of these packages have different ways to display geo, some correct and some not. So the people get used to using it, but not learning how polys covert to these different display level in render packages. Maya is a good package to learn on as most of its tools are aimed at the most basic poly construction level. Not saying maya is basic, just that the tool are pretty plain and simple to use. maya is then left open for you to create more complex tool as your skills grow. Therefore if you can create a awesome model using a package that has simplistic tools then you will be able to awesome model in a package that overlooks some of the basic poly principles.

Hope that made sense. Kinda does in my head. :thumbsup:

[El Burritoh]

John,

I think that makes some sense in my head too, although I can't say I understand you fully. Perhaps you can elaborate? I definitely know what you mean about how models created in one package can look different in another. I see it primarily in whatever realtime shader is being used. I don't see it so much at render time. I know XSI shows me more accurate smoothing than modo does once I move my models across, but some of that is also due to XSI's Geometry Approximation settings. Anyway, yes, I have run into that, so I try to be careful. I figure that if it renders well, it's good to go.

Being more of an XSI guy and working in an XSI studio, I haven't "needed" to learn Maya yet. But two things have pushed me to get my hands dirty with it lately. First, the fact that it's so extensible and open. It trumps XSI and Max there. And secondly, the fact that it's such a broad standard in the industry (due no doubt to the first reason). To be honest, I have always found the interface inelegant (especially hypershade and the node view! Can we please update that! It's like something from 1993), but I also know that it's just a question of learning how things are done. I had to learn some Max this past year to wear a FumeFX hat. I'll get the hang of Maya. The more you know, the easier it is to keep learning. I have a feeling Marking Menus are going to be my best friends, like they are in modo.



Ianucci,

That's a tough piece to model in SubD. I'm starting to wonder if you wouldn't be better off in Nurbs, or possibly even in straight polies, with a really high count. I don't think you can get that corner to look right without a much higher resolution in SubD. But again, it's really about how well it renders, and that brings us back to what John was just talking about. At least I think it does...

If you stick with SubDs, moving verts around will be necessary, but you'll also need extra density in some areas to get the right control you need. I don't have an elegant SubD solution for this. I would just try different things and expect to have to remodel it a couple of times.

[JohnSG]

El Burritoh

Yeah i think you get what i am trying to say, And by no means am i saying this is the word of god, just my opinion :D

In my eyes Maya and Xsi are very similar and at the same level for, as is Max but pushed more towards games. But all these packages can be used in a full pipeline. Modeling, Textures, Rigging, Animation, Dynamics, Lighting etc... so they are all very broad and usually cover all the basic areas well and leave it to the user to expand. While modo, (not that i have used it in a while, and i am aware it has more features then it did) is a hardcore modeling program, and in the right hand can be deadly. But before wielding it like a ninja i feel that some people have skipped over the basics of how geometry works and why it does what it does. Instead of letting a new program just smooth your geo in whatever way it can and make it look pretty.

For example the WIP car i posted, alot of the main feedback was about how little geo I was using to create all the shapes. Being able to understand how the basic geometry works and changes in what ever render package you use, polies, nurbs, subds. will give you greater control in these other packages.

1096

I don't use any of them myself, I find the basic tools in Maya just fine. understanding the geo that i use has helped me to view a piece of artwork or a onset model and know how i will model it, where the flow will be and what is a tricky pieces and what isnt.

Another example would be the hard surface video that was posted on the manga style power suit a page or two ago, It looked very impressive, but if you took that geo as an obj into any pipeline to texture, rig, animate or render you would need a few days to clean it up. Depending on your pipeline of course) As the software is calculating alot of the cuts and edits you are doing to the underling geo not you.

Basically walk before you can run, just cos a new package comes out with fancy tool and shiny buttons doesn't mean its any better or that it will make you better, know whats going on underneath as this is what will make you a better modeler. (info mainly for the new guys who just starting out in 3d):thumbsup:

[El Burritoh]

Yeah, I think I'm right there with you, John. When I first started modeling I think I had a pretty narrow understanding of what geometry really was, being limited to whatever package I had and not realizing the common science behind it all. I've had to work in enough apps since then to understand that a triangle in one is the same in another. My opinion is that where SubDs are concerned, as long as you're using Catmull-Clark, you can get pretty predictable subdivision results across packages, since the subdivision algorithms are the basically same, just trademarked differently. Unless I'm wrong in my understanding of Catmull-Clark subdivs.

As for modo, I far prefer modeling in it than in XSI, which just doesn't give me the precision I need. I've found that for many projects, the precision I want is overkill, but being able to construct a model as precisely as modo lets me is still important. From the little I've used Maya, it seems to share some of that with modo, perhaps just enough to make the transition easier for me. I wish that modo was so much more than it is, but you pretty much nailed it: it's still just a modeling app on steroids (with a few bonuses though). Thankfully, it fits nicely into our pipeline.

So how would you tackle Ianucci's problem, if it had been handed to you instead?

[JohnSG]

Due to the helmet being almost totally curved everywhere, any area that has a tight angle like Ianucci's problem will require higher level of resolution.

You can either make the entire helemat one piece but at a higher resolution as to maintain the curved surface once you include the acute angle.


or split the helmet into a couple of different section, the black and red section. This way you will be able to have the black area of the helmet at a lower resolution all around the helmet. While the red section will be at a high enough ros that it will maintain the surface curve at the acute angle.


I would go with the second option personally. Gives you better control over the different sections and also gives a clean break for uvs. making your overall job easier. :D

[ianucci]

El Burritoh: The model is a sub divided smoothed poly model, there are no tru SubDs in Cinema nor NURBS which I have never ventured to study. You stick the model under a modifier object, kind of like Max I guess. I presume that what you meant by 'straight polys?'

John: When talking about the differences between programs displaying subdivided models are you talking about viewport display or smoothing algorithms, both or something else entirely?

Cinema uses Catmull-Clark but also has options for 'Linear' and 'Catmull-Clark (ngons)' which handles ngons in a different way and is the default algorithm. My mesh has a couple of ngons but not in that corner area.

I heard Modo subdivides stuff differently so I dunno if that is using something other that Catmull-Clark.

I have used Maya now and then and always found it frustrating mainly because I think I've been so spoilt by the Cinema interface and have been using that software for many years. I'm also not interested in scripting which seems to be important to get the most out of maya. However I really need to get more comfortable with one of the big three purely to make myself more employable. I'm looking forward to trying out 2011 because of the interface overhaul.

[JohnSG]

John: When talking about the differences between programs displaying subdivided models are you talking about viewport display or smoothing algorithms, both or something else entirely?

I heard Modo subdivides stuff differently so I dunno if that is using something other that Catmull-Clark.

I have used Maya now and then and always found it frustrating mainly because I think I've been so spoilt by the Cinema interface and have been using that software for many years.


All Catmull-Clark should be the same, i don't know enough about the other packages to really go into any detail.

I think what I was saying is similar to what you said about how you are used to Cinema and its tools", and being so used to it can sometime people get to lock into a certain way of doing something and this can effect there ability to advance or grow as a modeller. I tend not to rely on fancy tools that much which kinda forces you to understand the basic tools very well. I hope you do get a chance to look into using maya someday.

[El Burritoh]

Concerning modo's Subdivisions: Modo has a subd preview like Lightwave had (hit the Tab key). However, this preview mode is not actually a Catmull-Clark mode, but an extremely close approximation of it. To subdivide your mesh using CC, you have to physically subdivide your model, which then implements full CC. And since modo doesn't have "live" SubD modes or construction history, this operation is a one-way street.

In theory, this really stinks. In practice, it has only caused issues for me about 5% of the time at most. And most people move models from modo to Max, Maya and XSI all day and never notice anything. Modo's implementation of pseudo-CC for its realtime preview is actually very accurate. Although many of us wish they'd give us the real deal anyway, just to have our bases covered.

I don't model in SubD anyway, unless I want to check proportions. I do the vast majority of my work on non-subdivided polygons. It's not hard to predict how a mesh will subdivide, which may be why I don't have many issues moving geometry around. The tension between two given points will be the same in any final renderer that's likely to be used.

On another note, my NDA should be up in a month or so, once the product releases, so maybe I'll finally have something interesting to post here! It's something final, but I've wished so many times that I could get feedback here first. Stupid NDAs. My work suffers as a result if you ask me.

[mynameisnick24]

For example the WIP car i posted, alot of the main feedback was about how little geo I was using to create all the shapes.


So then, Which of the following would be the best choice in terms of amount of geometry?

1605
1606
1607

Initially I would of picked Example A, because I was always under the impression that I needed to make all of the polygons as even in size as possible throughout the mesh.

But now I'm thinking something like Example C would work just fine.

Is this correct, or is it a bit more complicated than that?

[weiluntsai]

for me, I would take example C style if there are so many objects in one project because C still keep the edge details and smooth edge but less poly count.
Just my opinion, hope to hear other master's comment:sw:

[tingles]

Example C for me:)
Especially if it's curvy ... you want to use the smallest amount of polys to get your curve ( depends on your object ) ... for example a car windshield is good to use less polys.
Could be wrong, but it works best for me !:)

[JohnSG]

I would also have to say C.

Heres an example i did.

This sort of method is good if you use a render engine that converts to subds at render time. It makes the poly count low and makes editing the geo quick and easy.

1613

1610

1612

If you do not use a render engine that converts to subds then this method is still not far off how I would approach a poly model.

In the crudest terms what subds do is smooth themselves, so if you have all your geo set out nice and lite, making it faster to edit and quicker to do uvs. Then at the end all you need to do to remove the low poly look is smooth it.

1611

Personally i never use smooth, I try and reduce the amount of geo at all times while remaining in control of the overall shape. But to each his own :thumbsup:

[El Burritoh]

I would also pick C, but sometimes I have had similar geometry as C give me trouble when adjusting UVs, due to the lack of geometry and therefore the lack of control points in the UV. I generally go with a lean version like C, but on occasion I'll add extra edge loops in there to make sure the UVs behave and don't go warping on me.

[ianucci]

One occasion where you might want something like A is if you are taking the mesh into a sculpting app in which case having evenly distributed polys is important.

[mynameisnick24]

Awesome, Thanks for the info guys. I will definitely stop adding all of the unnecessary geometry to my work, as most of my stuff was very dense like in Example A. On top of that I was smoothing things as well so, things got very heavy, very fast.

[andreicirdu]

here is a maya script for making 8 sided circles: http://www.creativecrash.com/maya/downloads/scripts-plugins/polygon/c/tdarrangecircle

What you have to do:

1. Pick a vert

http://www.cgfeedback.com/cgfeedback/attachment.php?attachmentid=3844&stc=1&d=1291125229

2.Run the script
http://www.cgfeedback.com/cgfeedback/attachment.php?attachmentid=3845&stc=1&d=1291125177

[JohnSG]

Very Nice, :D

[Kel Solaar]

Awesome, Thanks for the info guys. I will definitely stop adding all of the unnecessary geometry to my work, as most of my stuff was very dense like in Example A. On top of that I was smoothing things as well so, things got very heavy, very fast.

I think it all depend what you try to do, if you plan doing a lot of zbrush post modeling to add details, break the surface, etc ... you are better using the Example A because you'll have a better face repartition. With Example C you would have a lot of polys on the edges but maybe not enough at the center. In the end it's really what you wanna do that dictates which one is the better.

KS

[Time2Time]

Totally agree with Thomas! Always depends of what you need.

[Octo]

Just wanted to mention the 3dsmax plugin edgechex, which chamfers edges based on angle or smoothing groups, and its a modifier so it's extremely easy to work with..you can always go down to the base mesh and do whatever changes needed.

http://www.mattiasmagito.com/misc/edgechex.jpg

This example is very simple of course...but it works on just about anything I've ran into so far. It's also capable of a '3 edge' chamfer, but that one doesn't seem to work as nicely in all situations so you can sharpen up the chamfers like i did in the example. I've used this plugin since i discovered it and have gotten very spoiled...almost wouldn't work without it nowadays because it saves so much time heh.
It does create ngons here and there, and triangles..which can cause problems in curved areas...but then you can just solve it manually and set the whole area to the same smoothgroup...and imo the result is pretty damn clean either way..
anyways a more advanced example here, although it was a work in progress pic so the result isn't perfect

http://www.mattiasmagito.com/images/lotr/sauron_wip3_web.jpg

....laters! :)

[andreicirdu]

here's a maya python script that allows you to select similar faces http://www.creativecrash.com/maya/downloads/scripts-plugins/modeling/c/select-identical-components

How it worx:
1.put the script into your maya scripts folder
2.select a face
http://www.cgfeedback.com/cgfeedback/attachment.php?attachmentid=8551&stc=1&d=1314912117
3. run this python command:
import SelectIdentical
SelectIdentical.run()

4. here's the result:
http://www.cgfeedback.com/cgfeedback/attachment.php?attachmentid=8552&stc=1&d=1314912117