Sculpting workflow for full body figures...

[Djordr]

Hey I have sculpted two posed quite realistic figures, this is how I do it, I start with a sphere extrude the body from it head etc. Increase the dynamesh and add a few more detailed points to the model. Then I pose it still in dynamesh mode because I find z remeshing the figure and posing it often breaks the model too much.

So I put the figure in a good pose, Head, body and hair seperate. Start working assymetrically and then when the entire model is about 1.2 million polygons high. I come to a point where I want to add really fine details or slightly adjust the proportions of something and it overwhelms me, I WANT TO USE A SUBDIVIDED MESH INSTEAD. Sooo I guess I’ll have to project the detail from the dynamesh back onto a low polygon version of the dynamesh.

I make a copy of the body and face that at this point has to be merged with the body because it’s too late after you start subdividing the model. I do the projection work and about at 1.6 million polygons it looks like the original model…almost. Because the z remesher didn’t do a very well job getting the symmetry of the assymetrical model very well. So on the face for instance the area around the eyelids look funky as do other parts of the model.

To be honest the dynameshed model looks better but is far too hard to add super tiny details to. Is there any good workflow for making realistic and start working at them from a posed position? I guess good topology is crucial but I hate retopo work, the z rememsher doesn’t do a great job with it and going into an external program for something that is not a game model pisses me off. Is there any solution to this other than to make super perfect topology low polygon models to pose before doing the sculpting?

In an ideal world I’d like to go from, sphere > Rough human figure > Put into a pose > turn into a nice subdivided model to work with. But the symmetry of the human body while posed breaks, or it did for me anyway. Seems clunky, some good tips for dealing with this? It’s sort of a huge bump in the road when it comes to digital sculpting.

[Spyndel]

Nobody enjoys discipline, but it’s an important part of any efficient workflow. In this case you’ve discovered that dynamesh tends to work best with meshes at medium to medium high resolution. You’ve discovered it would be better to use dynamesh in early to intermediate stages to establish form, and then switch to a multiple subd level workflow to start adding fine detail. You’ve discovered that it would be better to pose a figure last of all, because it breaks symmetry (Poseable Symmetrycan sometime be of use, but it would be better not to rely on it as a primary option).

I say follow your instincts, and trust the things you’ve discovered through trial and error.

Zbrush gives you the freedom to violate this order, to situationally dynamesh a multi subd level mesh if you deem it necessary, but it’s going to require the use of tools like projection, and it’s going to cost time. You want to establish a workflow where you minimize the need for it.

[Djordr]

I looked at how scott eaton did it in one of his tutorials, he solved it by making a base mesh with really nice topology that would deform well when posed. It’s just sad though that you pretty much have to discard dynamesh all together for that and only work with subdivisions, I mean I managed to make this, just from a sphere.

For there to be no good way to turning this into a subdivided model that it would be easier to work with, it’s quite a nasty workflow issue. Instead you have to make a perfectly symmetrical human model, with perfect topology for posing without deformations. I mean most of the sculpting would take place during the symmetrical phase.

I pretty much have to abandon this at this point, because it’s too hard to work with effectively, huge turn off for someone that is not too interested in animation or game models or perfect topology flow.

[Spyndel]

No, you don’t have to work one way or another. My original post was from more the point of view of a production workflow. If you want to work in a more sculptorly–is that a word?–more like a sculptor in the real world, you would just shape your clay (dynamesh) into the rough pose you want, ZRemesh for better low poly topo, subdivide and start sculpting detail. It means depriving yourself of some digital “cheats” like automatic symmetry , but a real world sculptor wouldn’t have that, nor would they want a figure that looked too perfectly symmetrical. Based on the Scott Eaton tutorialIve seen, he is a confident sculptor, and abandons working with symmetry after posing a rough symmetrical figure.

If you adopt a workflow like that, you can make use of the Poseable Symmetry function I linked you in my previous post.

And yes, you can turn that model into a mesh with subdivision levels at any time. The default ZRemehser settings should give you adequate polys for a clean well dsitributed topo low poly version, which you can then subdivide as desired and project the detail from the original mesh onto it with the procedure I linked in my previous post. You can repeat this as many times as necessary, if you ever want to dynamesh something again, or remesh topology that has become distorted.

If you just want to stay entirely in dynamesh, you can try increasing the resolution slider to capture more detail, or increase the size of your model which will increase the amount of resolution devoted to it. Just keep in mind Dynamesh has a practical upper limit for detail, and eventually your machine will struggle to dynamesh very high resolution meshes.

[Djordr]

Sculpting without symmetry, more work but a good habit maybe. I’ll try z remesh the model again, and then add the detail back again to the face and areas that have been distorted, assymetrically. Even using assymetrical z remesher guides on the face is doable I guess. Thanks.

[Spyndel]

If you don’t mind my asking, why are you bothering with Zremesher guides at all? You said you didn’t have much interest in animation, and to this point you’ve been working with dynamesh’s somewhat erratic topology, so your need for highly optimized deliberate topology would appear to be minimal.

All you really need is fairly clean, even quad distribution for the best sculpting performance, and enough poly resolution to capture the structures on the face/body (nose, ears, digits etc). The default ZRemesher settings should be sufficient if you you adjust the target poly count until you get the base detail you want, and maybe bump the adaptive size slider up a bit if you want a little more accuracy at the expense of poly shape.

Remember you’ll get the higher level detail back with subdivision and projection, so don’t get too hung up on base level topo for your purposes.

[Djordr]

Good point, although some guides around areas such as nostrils and eyelids couldn’t be too bad. It’s mostly for areas that are tiny and need lots of detail, you don’t do eyelids with subdivisions unless you’ve got a really good computer I guess. But I’m on a laptop that starts lagging at around 3 million polygons. Anyway I appreciate your advice, thanks.

[Spyndel]

Ah, that makes sense. I suppose a hardware limitation could influence this. I just assume that nobody is working with a more meager system than my aging rig.

One last tip. You can use the “use polypaint” option in the Zremesher menu to simply paint areas you’d like more poly density in, and away from areas where you dont. This is generally easier than the trial and error that comes with guide curves, but in some cases may cause somewhat distorted topology. Your mileage may vary.