ZBrushCentral

Annoying divide mesh problem

Hello everybody

At the moment I’m adding details to a imported head .obj model in ZBrush. In this case it’s the ear. For control I increased the subdivision level and decreased again to keep on pushing an pulling at a low level polygon stage. As the work progressed I started wondering where some already worked out details have gone. :cry:

I made a test --> see pictures below: Mesh detail from original Geometry as imported:[img]http://www.zbrush.de/images/uploaded4/user_img-dPpt9JWqdw_1_before_subdiv.jpg[/img] please mind the marked Vertex! After dividing/setting subdivision level to 2 the result is as follows: [img]http://www.zbrush.de/images/uploaded4/user_img-BneU8ottTY_2_subdivided.jpg[/img] Setting the subdiv level back to 1 the result: [img]http://www.zbrush.de/images/uploaded4/user_img-9I8rDtCvV4_3_after_subdeviding_level1.jpg[/img]

Shouldn’t the marked vertex be at the same Position as in the first picture? Actually the whole mesh has been smoothed. What’s the reason for this resp. how can I avoid ZBrush to smooth/shrink the whole mesh when dividing? Please load for comparison the pictures 1 and 3 into your favorite picture browser and switch between those two pictures -the loss of detail is not too marginal…

Thanks for help in advance

kelvin273

the first picture is probably the original model you brought into Zbrush. When you smooth it and goback down to level one, the model is different. This is very normal in Zbrush. Zbrush is giving you the option to use the new model it created in picture 3 or your original model in picture 1, for displacement map generation. The Zbrush model has the advantage that it will always more closely resemble your hiRez model. If you want to use the original model you first brought in, just go down to level 1 and reimport it back in, then when you generate a displacement map, it will be based on the original model

Thanks for your help Long Phan.
Yes, the first Picture shows the mesh right after importing into ZBrush and the last one shows the same model after dividing and setting the subdivision level back to 1.

Now I understand a little better why ZBrush is working this way although this smoothing behavior makes no real sense to me. I really can see no advantage in this ‘feature’ of ZBrush. Far from it!
This means that it is impossible to keep a 3D-Geometry quite accurate while working in ZBrush and if I’m adding some detail in level one and afterwards divide the mesh for adding more detail in higher resolution, I will always loose a lot of them so that I’ve to re-do them again in the higher resolution?
And also: When would there be a case that it is desired to apply a displacement map on to a different model/geometry than the original one that was imported?

Greetings kelvin273

EDIT: for explanation what I mean with accurate modeling and what ZBrush’s smoothing causes, I’ve put reference geometry over the mesh that was partly modified (nose area) in ZBrush and exported again as .obj into another modeling software.

…thanks to ZBrush I’have to do now a lot of modeling to get the original head shape back :mad:…

Multi-resolution subdivision editing means that when you make any changes to any level of the mesh and then switch to another level, those changes will be interpolated across all other subdivision levels. This applies both up the levels and down.

Example: You are sculpting a croc’s head and decide that you want the mouth to be longer. You go to level 1 where you can move just a few points to make the mouth longer. When you return to the highest level, the teeth and all other high level details will have moved to fit the changes made at the lower level.

The same works in reverse. When you make changes at a high level, the lower levels of the mesh will be adjusted to fit.

Example: You are sculpting a head and decide that you wish to add short devil-type horns. You use Projection Master and an imported head mesh to add a face to the pommel of a sword. When you move down in subdivision levels, the lower-level geometry has been updated to come as close as possible to the shape created with Projection Master. You can then continue to refine that shape if you wish at the lower level before returning to the high level for more detailing.

See how it’s advantageous for the Multi-Res features to work both ways?

Now to your “problem”. ZBrush uses the catmull-clark formula for subdividing a mesh. This includes the mesh smoothing algorithms. Why does it use this method? Because most animation software does, too. By using the same technique, it ensures that the details you create via displacement mapping will match the ZBrush version of the model.

Well, the thing about mesh smoothing using this algorithm is that the mesh will contract when it’s applied. And details get smoothed out. Take your same model and using your animation package apply a hyper-NURBS to it. What happens? Does that detail that you pointed out in your illustrations disappear? Yep! Can you superimpose the smoothed mesh over the unsmoothed one and get the same result as in the last image you posted? Yep! It’s the way that the math works. Try it once with a six-faced cube to really see some dramatic differences. :wink:

Now, because Multi-Resolution Subdivision Editing causes all levels to be updated when you switch between them, level 1 will invariably be modified by the smoothing that takes place as you divide the mesh. This creates zero problems, however, because ZBrush also lets you import that original mesh back into level 1. By doing this right before you create the displacement map, the map is able to be calculated for the exact mesh that you are animating with, while still enjoying the many benefits of Multi-Res Editing up to that point.

Hope that helps!

Thank you aurick for this useful information! When it comes to trouble you really stand out to be the nuts and bolts of this board.
:+1:small_orange_diamond:+1:small_orange_diamond:+1: Tanks alot for your dedication! :+1:small_orange_diamond:+1:small_orange_diamond:+1:

So from now on, when I do some changes in ZBrush I will instantly reload the model a couple of times after doing some dividing in order to keep the smooth-altering of the model down to an acceptable degree. Unfortunately I never came across this problem before and I think this issue should be pointed out in the manual very clearly.

As a workaround for my model I load the modified Model into silo in addition to a previous (correct topology) version and will try to cut and scale,patchwork,puzzle and merge some parts of the polymeshes together. Maybe I afterwards will do some move tool session in ZBrush with the masked reference mesh in the back. Let’s see…

Greetings
kelvin273

EDIT: Also found a useful info Here
Now I know what’s the cage button for :bulb:

Hello again.
I’m having annoying troubles with this dividing mesh again :frowning:
When working in preview adaptive skin mode according the Pose2Pose technique, what can I do to protect a mesh from being distortet ? Everytime I push the slider my Mesh is undesiredly distroyed again and again. In Preview mode it is impossible to re-Import the mesh for keeping the original shape. Also the cage funktion is no real help because it works very inaccurate and I have to redo lots of mesh parts again and again every time I save the model. Because of the model then switches to ZSphere mode and smoothes into highest level (in my case SDiv6) when pressing ‘A’ again. If I then get back to a lower res mesh the model is bad again

Can anybody help me with this problem or knows a workaround?

Thanks in advance
kelvin273

It would help tremendously if you actually described the exact steps that you’re following, and showed examples.

Sorry Aurick
Here are some Pictures.
This is the original Mesh as it comes from the ZSpheres into adaptive skin preview
[attach=13635]01_Mesh_orig.jpg[/attach]

This one shows the mesh after Vertex moving (Adaptive Skin res=6, Geometry SDiv=2) [attach=13636]02_Mesh_ASkin6_deform_geomSDiv2.jpg[/attach] This is after storing Morph Target and pressing 'A',Geometry SDiv=2 and 'Cage'-button [attach=13640]05_Mesh_ASkin6_deform_geomSDiv2_cage.jpg[/attach] This one is the mesh (after doing cage in SDiv=2) and put slider to Geometry SDiv=6 :( [attach=13645]051_Mesh_ASkin6_deform_geomSDiv6.jpg[/attach] This is the Model after Posing :confused: [attach=13641]06_Mesh_ASkin6_deform_geomSDiv2_posed.jpg[/attach] And in Higher res (ASkin=6 SDiv=6) :cry: [attach=13642]07_Mesh_ASkin6_deform_geomSDiv2_posed.jpg[/attach] So there are 2 Problems at once: 1.)How can I get the Mesh to be bound to the ZSpheres for posing and 2.) not to be distorted in higher resolution?

Thanks for Help in advance
kelvin273

Attachments

01_Mesh_orig.jpg

02_Mesh_ASkin6_deform_geomSDiv2.jpg

05_Mesh_ASkin6_deform_geomSDiv2_cage.jpg

06_Mesh_ASkin6_deform_geomSDiv2_posed.jpg

07_Mesh_ASkin6_deform_geomSDiv2_posed.jpg

051_Mesh_ASkin6_deform_geomSDiv6.jpg

First off, don’t use the Cage button unless you’re ready to create a displacement map and export both it and the base mesh. The cage calculates a mesh that as closely as possible resembles the high resolution mesh. It is meant for exporting to another program; not for use in ZBrush itself.

What you want to do is the following:

  1. Create your ZSphere model in the default pose.
  2. With the ZSpheres visible, press Tool>Morph Target>StoreMT
  3. Turn on the preview and assign any UV mapping that you want. Also do your sculpting.
  4. Return to the ZSphere view and repose the model.
  5. Activate the preview to see the new pose.

The morph target is stored before you ever start sculpting or texturing (at any subdivision level, including level 1) and the cage button is never used.

Thanks again for trying to help aurick.
I tried the procedure as you have explaned above, but it just doesn’t work for my model. I always get that bad distortion in the wings when posing the model as shown in the last 2 pictures of my last post (aside from the Problem of strong smoothing/loss of shape in the membranes in higher SubD-levels).
I guess this posing feature doesn’t work if some vertices are moved too far from the ZSphere. :frowning:

For anybody to tryout attached you find the model. Maybe there is just something important I missed that the model always comes with bad results?

Regards kelvin