Zbrush to Second Life Workflow

[pepper Mocha]

I've been researching Zbrush to SL workflow but I see all the tutorial information i'm coming across is for the most part outdated.

Is there anyone who could help me (in basic terms) to list what the current steps are now for creating a mesh in Zbrush and taking it into Second Life (or) direct me to any current tutorials?

Programs I use/have-

Zbrush: sculpting

Blender: converting/weighting

Photoshop: texturing

 

Thanks

[arton Rotaru]

• Sculpt high poly model in ZBrush.

• Import high poly into Blender

• Retopo the high poly to make a game ready low poly model

• UV map the low poly model

• Baking high poly to low poly normal map, ambient occlusion map

• Texture low poly model

• Rigging/Skinning the low poly if necessary

• Create LOD models from the low poly, and collision meshes

• Import into second life

That's the basic workflow.

[pepper Mocha]

Thank you arton, your list is helpful.

It seems like all the other steps, except sculpting,are done in Blender? Do most people avoid making maps in Zbrush? And can texturing be done in Photoshop? Is this just a matter of preference or are there certain reasons, like Zbrush isnt as good for those steps?

Also, I'm not familiar with a collision mesh- is that optional or dependent on what your making? In my case I'm doing mesh head and hairs, would I skip this step as well as the LOD models?

 

Thanks

[arton Rotaru]

Well, that's just the basic steps. You can use all your creativity, and imagination to achieve the results that you are after. You can indeed paint that high poly sculpt as well in Zbrush. And bake that color information onto the low poly mesh as well. How practical that is may depend on your personal preferences. Besides sculpting Zbrush is great for creating "just" textures, like brick/stone walls and the like, where you only extract texture maps, but don't use the mesh itself in-game.

Personally I do most texture work on the low poly mesh directly though.

The important part of that workflow list is, that you don't use a sculpted high poly mesh straight out of ZBrush as the in-game model. Even if it's decimated, or ran through ZRemesher. It won't be a game optimized asset in most cases. You can achieve much better results with manually retopo-ing the high poly. So yeah, most of the low poly work is done in other programs.

A collision mesh, aka physics shape, is a very simplified representation of the model, which will be used by the physics engine, to calculate collisions on the object. For avatar attachments it would be enough to upload a simple cube along with the mesh, to keep upload costs low. Attachments do not collide, so their physics shape is less important.

LODs though are quite important to keep the render load balanced, and maintaining a decent look from any distance at the same time. So it's best to create your own LOD models as well, to have ultimate control over the visual appearance of your meshes. With organic shapes it might be enough to make usage of the auto LOD generator in the SL importer. But I wouldn't recommend that. Because a human can do much better, and I don't want to have a machine, more or less, destroying my work I had put so much time and effort into.

An example of neglecting the proper creation of LOD models with mesh hair, can be seen here. The user is asking why her hair went missing when zooming out.

[say Moo]

with 4r7 release of Zbrush, the zremesher (auto retopology tool, aka high to low poly process) is very good.. (you can draw curves to follow, and density areas that need to be followed during retopo). Also Uw mapping (via uv master plugin) is now rather good. Same for the baking process (normals, specular, diffuse, ao) No need for a traditional modeler for those (although you are ofcourse free to use those for these steps, if you need so)

For animation, rigging  (non static mesh) though.. it is still needed. (no good way/support in zbrush)

[Nacy Nightfire]

One approach is to sculpt a dense detailed model in Zbrush from a densly divided polymesh 3d "tool" of your choice (now made easier with dynamesh).  Then you retopologize the dense mesh* either using zsphere re-topology  or by  creating  automatic retopology with  ZRemesher using control curves.  Use projection in the subtools menu to transfer the detail from the original  sculpted model  onto the retopoloogy mesh.  Add additional  subdivision levels to the retopologized  mesh and contunue to reproject the details at each level of subdivision.  Drop the retopolgized mesh down to its lowest subdivision level to UV unwrap with UV Master .  You will also have to set your retopologized mesh to the lowest subdivision level in order to create a normal map in Zbrush.  All painting/texturing is done at the at the highest subdivision level and that "polypaint" is then transferred to a   texture which will be applied to the exported lo subdivision level mesh.  Export the lowest or close to the lowest subdivision level.  Import this low poly version of the mesh into Blender to rig/skin and to create lods (or you can make your lod meshes zbrush, but it's probably easier to do this using Blender),  export the mesh as a collada file.  Export out any texture, normal and or displacement textures you make in Zbrush for uploading to Second Life or to bring into Blender and/or Photoshop for further baking and texturing.

I've used the above method but it's not the one I prefer.  I prefer to create a base mesh in Blender with the correct edge flow and import that base mesh into Zbrush for sculpting and I add subdivision levels as I add detail when sculpting. This workflow does not require retopology after the fact.   I find it takes less time to think the project through at the beginning and start out with the edgeflow that works best for the mesh I'm going to be sculpting then it does to retopologize after sculpting. 

Usually I go back and forth between Blender and Zbrush, I might texture in either program or in photoshop or do a bit of texturing  in all three programs.  For the most part I frequently complete the mesh in Blender then go into Zbrush just  to give the mesh a few tweaks to correct proportions, etc.

I think you can spend a lot of time trying to force Zremesher to do your bidding, but it's just easier to retopologize with zspheres if you are creating topology for characters that will be animated.  Edgeflow is really critical in that case.  For static items then for the most part the results you get from Zremesher are generally acceptable.

*I don't recommend that you try to import that initial dense sculpted mesh into Blender unless you have an extremely powerful computer.  You plan should be to import the low poly retopologized version into Blender.  The whole point of using zbrush is it's uinquely designed to allow for sculpting super high poly meshes which is not possible or practical in standard 3d programs.  This is why it us designed with a  single viewport and doesn't allow for multiply views of the object (quad view).