ZBrushCentral

Zbrush pipeline and re-topology woes

Hi Zbrushers,

I have a couple questions regarding workflow with Zbrush. I have been using it enough that I’m comfortable sculpting in zbrush now, and have a good flow between Maya, Modo (for my organic UV’s) and Zbrush. I’m trying to figure out a professional and reliable pipeline for my work, something that would make me a dependable freelance artist not green about things that have an answer. Roughly put my pipe goes something like this.

  1. sculpt basic shape of character in Maya.
  2. export to Modo where I apply UV’s (best organic for me)
  3. import back into Maya to maintain UV’s and export as obj to Zbrush
  4. sculpt my character out in Zbrush and export new Low model, normal map and displacement map (32 bit) to Maya.
  5. render using mental ray and there is my work, although it is not flawless. I have problems with displacement maps and even normal maps that don’t resolve accurately in Zbrush.

I’m trying to resolve these issues and tighten the workflow on my pipeline, and take my work to the next level. Toward this end I have been reading a lot, and watching a lot of demos on Zbrush workflow. This of course has lead me to see that artists talk about retopologizing their zbrush sculpts when they bring them back to 3d software which makes me do a double take.

Question 1)
Does every artist re-topologize back in the 3d software if they are working for games or film or basically anything that moves? This seems like a painful ending to using software that is meant to be intuitive and improve work flow. Am I looking at dated material (gnomon tutorials I purchased this week, Feb 22 2009) Character Modeling for Production.

Question 2)
I get my normal map, and displacement to work in Maya however when just using Normal map (like in games) the character lacks the volume of the displaced character. If gamers are only using normal maps and not displacement maps then how do they accomplish this volume? Is it because they are re-topologizing (i.e. rebuilding the model back in Maya or Max or 3d software of their choice?) That doesn’t seem like the right solution because even though my model is simple, it is around 6000 polys, and to re-topologize something complex from Zbrush would surely put me much higher, and take a load more time. It seems to defeat the purpose of using Zbrush to begin with to have to “trace” over the exported hi-model back in Maya or Max.

Question 3)
So why re-topologize? I can see from demos that they send hi-res zbrushed “sketches” back to Maya, and recreate them in 3d, but there is no discussion of why this is necessary, and I don’t know if they do it just because, A) the video is old, B) That’s all that studio knows how to do C) It’s because animated characters require movements that would stretch geom incorrectly rendering zbrushed models useless in production.

The thing is that I don’t hear people talking about this. Saying, Yes Zbrush is brilliant but you know you have to rebuild the model in 3d software if you actually want to do anything other than a still pose or illustration.

This to me seems like it could be a major drag, since I like to create and sculpt and not necessarily spend all this time learning a cool piece of software to find out that I have to yet again return to the painful path of pulling polys.

-Daev

Hey, there daevfinn@mac.com.

When it comes to the Zbrush pipeline it is all about what feels comfortable for you and what gets the job done. Personally my pipeline is:

  1. Make base model in Maya then export out OBJ.
  2. Do sculpting work in Zbrush.
  3. Export out displacement map, export a polypainted mesh (color map), and normal map. Also export out a base mesh from Zbrush.
  4. Tweak some of the maps in Photoshop.
  5. Render out the model within Maya using mental ray.

Those are just the general steps I take to render out a model out of Zbrush.

Personally I do most of my retopologizing within Maya because it gives me more control and freedom. Although there is a retopologizing feature in Zbrush that I have tried but never got the hang of it. I too have purchased a number of tutorials from Gnomon and they have helped me in some way or form. If you are looking to learn in depth about stuff in both Maya and Zbrush I would suggest Alex Alvarez’s tutorials. He does a pretty good job of explaining things.

Now about the Normal maps. You have export out a base mesh that has enough silhouette to represent your highest sud D level or else it is going to look flat in the renders. I would suggest exporting out maybe a sub D level 2 or 3 mesh out of Zbrush. Your final base mesh for normal maps should be around at least 25,000 polys total or more. Don’t go any higher than 80,000 or you will start to lag in Maya unless if you have a really powerful computer. The key thing to understand about Normal maps is you have export out a silhouette that best represents your highest sub D level. I would recommend you pick up the book “Zbrush Character Creation: Advanced Digital Sculpting” by Scott Spencer which explains a lot about displacement maps and normal maps.

Having bad topology within Zbrush will only cause more problems to your mesh when you start subdiving. Zbrush does not like triangles or any topology that has more than 4 sides meaning that if you have a face that has more than 4 sides then that will cause problems in Zbrush. Also retopoligizing also helps to deform creases more naturally where your arms and legs fold. In the areas where the arms and legs crease you would need more edge loops for it to move naturally.

As for going back to Maya, yes it is painful to go back to pulling verts. I know because I have done this so many times. One thing you can do to speed up your workflow is by setting up Hotkeys by going to Window>Setting/Preferences>Hotkey Editor. Within the Hotkey Editor you can set up hotkeys to the commands that you mostly use. I would also recommend that you use Convert Selection, you can find it by going to Select>Convert Selection. The Convert Selection really makes selecting easy and faster within Maya without ever clicking more than once.

All I can say is you have to experiment and do some trial error in order to find out what works and what doesn’t. I know I have because that is how I learned. It is one thing to watch someone else do it but actually doing it yourself is a totally different story. You have to be able to experience it in order to learn it. Search through forums on the internet or whatever you can get your hands to better understand how to work efficiently. Patience is key, it’ll come to you naturally once you get the hang of it. :smiley:

Hi Seraph769,

Thanks for the response and insights, it is helpful. To share a little of what I’ve learned on the side. I contacted a friend and former colleague from Rhythm and Hues. They recently used Zbrush to create the Hulk and Abomination. I expressed some of my same concerns here to discuss pipeline issues and they agreed that Zbrush is great but they are still getting the hang of it in production.

My friend recommended a Maya plugin called Nex which allows you to retopologize and have it magnet down to the surface of your mesh. I’m not positive magnet is the right word, but along those lines to give you an accurate retopology.

I went to find this software but unfortunately I currently only have Maya on my Mac, and this plugin does not work for the Mac OSX or maybe it works on Mac Intel, not sure. My system is robust but I bought it just before the Intel systems were released because I needed the system and didn’t want to be among the first wave of intel Mac users. Anyway, I think this is valuable information.

My Zbrush skills are definitely getting stronger, but I do think people need to be aware that what can be knocked out quickly as a sculpture in zbrush may have less real world application if people are not careful to think about a pipeline for UV’s and retopology, and yes strong models going in. Personally I tend to never use any of the default models from zbrush to work with, feeling that dulls my ability and understanding of the mesh. I think those lovely default models are a mixed blessing.

It’s just the reality right now that one software package cannot do it all.

-Daev

I gotta question. Does someone else within the studio usually do nothing BUT retopoligize models? Is there a separate artist that models and sculpts in hi-res and then passes his model to another guy who will retopo the model into low-res?

Our workflow (for games) is:

  1. Sculpt the character in ZBrush. Sometimes we’ll build bits and bobs in another app first to get a better starting shape, and we’ll often import a template or guide mesh of some sort to sit as an unedited (and temporary) subtool if the character has to fit a pre-existing rig, but often the whole character is created in ZBrush from scratch.
  2. Export a Decimated version of the finished sculpt to another app for retopologising and UVing (and rigging and exporting etc). And just to be clear, when I say “retopologising” I mean “rebuilding completely from scratch”.
  3. Export this fella and the original sculpt and bake them in XNormal.
  4. Paint the textures in PS.

On the odd occassion when we use polypaint inside ZB we just use auto UVing to get a texture out and bake it at the same time as everything else, but none of my team much like polypaint, myself included. I only use it if I know in advance that we need the high res version textured, for marketing renders or whatever.

I’m trying to resolve these issues and tighten the workflow on my pipeline, and take my work to the next level. Toward this end I have been reading a lot, and watching a lot of demos on Zbrush workflow. This of course has lead me to see that artists talk about retopologizing their zbrush sculpts when they bring them back to 3d software which makes me do a double take.

Question 1)
Does every artist re-topologize back in the 3d software if they are working for games or film or basically anything that moves? This seems like a painful ending to using software that is meant to be intuitive and improve work flow. Am I looking at dated material (gnomon tutorials I purchased this week, Feb 22 2009) Character Modeling for Production.

All of my team do. It is tedious, but it’s unavoidable.

Question 2)
I get my normal map, and displacement to work in Maya however when just using Normal map (like in games) the character lacks the volume of the displaced character. If gamers are only using normal maps and not displacement maps then how do they accomplish this volume? Is it because they are re-topologizing (i.e. rebuilding the model back in Maya or Max or 3d software of their choice?) That doesn’t seem like the right solution because even though my model is simple, it is around 6000 polys, and to re-topologize something complex from Zbrush would surely put me much higher, and take a load more time. It seems to defeat the purpose of using Zbrush to begin with to have to “trace” over the exported hi-model back in Maya or Max.
I don’t really understand what you mean. The purpose of using ZBrush is to create a super-high-detail sculpt that can be baked onto a low detail real-time model. What purpose is defeated?

Question 3)
So why re-topologize? I can see from demos that they send hi-res zbrushed “sketches” back to Maya, and recreate them in 3d, but there is no discussion of why this is necessary, and I don’t know if they do it just because, A) the video is old, B) That’s all that studio knows how to do C) It’s because animated characters require movements that would stretch geom incorrectly rendering zbrushed models useless in production.
Because a mesh that’s suitable for sculpting in ZB is completely different from a mesh that’s suitable for a low-poly character. A base sculpting mesh needs to be as regular as possible - all quads and as few poles as you can manage, and the distribution of detail is more important than its placement. A game mesh needs to be optimal - quads and poles are irrelevant and the detail needs to be distributed only to areas that affect the sihlouette. They’re two completely different things and the worst mistake an artist can make (IMO) is thinking of ZB as a “finishing” tool for adding detail to a pre-existing mesh. A game mesh simply will not subdivide in an optimal way for sculpting.

The thing is that I don’t hear people talking about this. Saying, Yes Zbrush is brilliant but you know you have to rebuild the model in 3d software if you actually want to do anything other than a still pose or illustration.

This to me seems like it could be a major drag, since I like to create and sculpt and not necessarily spend all this time learning a cool piece of software to find out that I have to yet again return to the painful path of pulling polys.

You’re looking at it back-to-front. We’ve always had to build low-poly models and ZBrush hasn’t changed that (or tried to). ZBrush has just allowed us to do much more with them.

Is there usually one guy who does nothing but retopologize or recreate a low res-model from a high res model? Or does one modeller have to do both?

In this studio an artist is provided with concept drawings and then creates the asset from start to finish (exported and working in-game), so the guy retopologising it would almost certainly be the same guy who sculpted it.