a Workflow Concept for 3DCoat

[michalis]

I'm wondering, all Farsthary's work on tools (when finished), is it possible to mirror it on sculpt room? (A MV workflow)

Another UI issue for me is, using MV method, paint and sculpt room should be one. These aren't bumps that we paint there, these are real displacements. Pinching on UV maps isn't right though, it deforms them.

I think that MV workflow in 3dcoat is completely wrong. I think that after retopo and UVs we should start seeing results on the real sculpting room (not vox room anymore). Then work on it and go directly for baking maps.

[polyxo]

I'm wondering, all Farsthary's work on tools (when finished), is it possible to mirror it on sculpt room? (A MV workflow)

Another UI issue for me is, using MV method, paint and sculpt room should be one. These aren't bumps that we paint there, these are real displacements.

In my opionion one should be able to go back into the Workspace which is currently called Voxel.

The Room which is currently called Sculpt should be removed. Having two - in fact even three Rooms where one can Deform the

Object is hard to justify as a concept and a source of struggle when learning the application.

I think one should be able to use all Mesh based ("Surface") tools directly on Stuff coming from the Paint-Workspace.

All transforms and Pose operations should work even with textures showing, mesh based deformation could work on top of the Retopo Mesh.

Those Tools which are Voxel-based or Mesh based but actually changing Topology (LiveClay) imo should be performed on the (Pre-Retopo) Input Voxel Object

and use a process which equals Project All in Zbrush - which transfers all Edits onto the Retopo-Object.

I believe in an elegant implementation of such one could eventually give up on the Room concept entirely.

PS: I still liked the option to Paint Colour on "Dense Quads" and to freely perform whatever further Sculpting on them at any point in time.

While I clearly understand the need to transfer dense Meshes to regular Low-Poly versions for all Animation geared work - such is typically not required when creating

Still Imagery (also if Michaelis is addicted to Retopo )

[polyxo]

Just as a play of thoughts:

Imagine you start a Sculpting Session with Voxels and Surface Tools.

Now - without Retopo and UV-Work - send the Model to the Paint Workspace as Dense Quads (3DC would use a refined implementation of PTEX for this

which gives you very nice resolution right away).

You paint Paint Texture and Bump until you realize that something larger has to get changed in the model.

No problem - you Switch back to Sculpting. 3DCoat displays the Voxel-Object again and you do whatever you like.

(It might be possible to have a viewmode showing the textured Dense Quads and the Voxel-Imput on top of each other.)

When finished with Voxel-Editing the Result was projected onto the already textured Dense Quads. Large topological changes like Holes

might be dealt with by merging painted Dense Quads with a new Dense Quad Version of the changed portions. The Ptex information of course had to be kept.

You finish the Texture of your Model.

Now you can either export directly and render with Ptex. Or - what any Animator will want to do:

Create a Retopo Mesh of suitable resolution, create UV's and bake Textures from the High Resolution Mesh. Or Retopo but skip the UV's and Bake to Ptex

but onto the Lower Res Version of the Mesh.

Ptex in this case would be used analog to the Vertex-Painting Zbrush uses - with all the Flexibility but the added advantage over the Zbrush-Method that one could

utilize its output directly (for Stills at least).

Thoughts?

[michalis]

Agreed polyxo +1.

Something like these.

Zbrush has a bad reputation speaking on UI issues. What most users don't understand is that zbrush has an effective UI with some hidden sometimes logic.

I believe that digital sculpting and texturing apps make possible lot of complicated and personalized methods. Really a PITA when thinking what UI is the most appropriate.

A good example could be this: A quality - hi definition sculpting of a posed hand.

Start sculpting, make fingers spread as possible- retopo it. Bake tex and rig it.

No, it won't work for quality sculpting. Now, after posing, we have to resculpt the surface. > zbrush like.

also if Michaelis is addicted to Retopo

LOL, how can I transfer some nice textures-displacements, even on static models polyxo? A remeshing, a projection on a low poly cage is needed. Clever loops may be for animators but some loops always needed.

BTW, just for static models, I developed a quassi retopo method using triangulated decimated cages. Super fast, effective and some times better than autopo.

Work in sculptris at 1M, export the obj, decimate in sculptris as possible, use even the remesher-manual reducer tool. Do UVs (using the clever hidden smooth operation) export this. In blender multires- subdivide and shrinkwrap. Bake hi frequency details on the lowdef cage Voilà! A quassi method, still the same principles.

I developed this workflow using 3dcoat when constructing this ancient like temple in 3dc gallery. Best decimated tri topology still comes from zbrush decimation-master though.

[polyxo]

Zbrush, regardless of what one may think of the GUI offers perfectly non linear workflows.

That's what I believe is one of the keys of its success.

One at any point in time can make large changes, even if one already spent a lot of time with Painting.

Nothing will be lost if at all possible.

3DCoat in its current incarnation is very linear. Its rooms equal pearls on a string.

Everything is has its clear position in a sequence of actions. Breaking out of this determination (by

exporting and re-importing into other rooms) requires quite a lot of understanding of how digital scultping works

under the hood.

However the Vertex-Paint approach of Zbrush also has disadvantages:

One can not Paint on Low Poly stuff.

By using a sort of refined default Ptex-workflow one might be able to get the best of the two words.

The same fuzz free flexibility while giving People options to paint on Lowpoly, Tris and Quads and Ngons

Also it might be an elegant way to get rid of the rigid Room system entirely.

LOL, how can I transfer some nice textures-displacements, even on static models polyxo?

Thing is that Displacement in its DNA is an Animation Technique.

It is not there in the first place to speed up rendering! It allows Animators to deal with Low Poly-Meshes when setting up Movement.

For this purpose Displacement makes most perfect sense.

If you are a Stills Rendering Guy you - as a matter of fact - do not need displacement in terms of Geometry-Handling-Improvement, right?

It's in my perception mostly a matter of optimization of Render-Engines. You often simply get sold the Workflows which are perfectly

valid for Animation as equally important for Still Imagery. Athough they actually arent (in principle at least).

Indeed due to their optimization Renderers therefore choke when letting a mesh of 500.000 faces render.

Maxwell however doesn't - it renders half a million faces just as fast as 10000 (without Displacement) Maybe because it was intended

as a Stills Renderer from the Beginning.

In such contexts I believe it's much easier to not care about Perfect Retopo and Displacement at all and instead use a relatively

High-Poly mesh+Normal-Map.

[michalis]

@polyxo

Interesting all these, we may be out of topic now. A bit.

The reason I'm using some kind of retopo is to be able to have a multi resolution model with UVs. I may decide to export a 300 K version with bumps baked from this point. I may need to avoid projection technics in blender, though I need a multi-res model there. One of the great great features of zbrush is this' re construct subdivisions", not in blender.

Another one about stills. Physically correct render engines don't always support normal maps. Another PITA now.

About vertex painting limitations: to paint on a low poly mesh isn't a problem normally, as you can subdivide up to millions first and then capture-bake the texture on UV map.

This "normally" stands for the disadvantage of this method, to paint on stretched faces, sharp tris etc. Better avoid to subdivide such faces.

Lot of ideas, lot of possibilities, in most cases they lead to a dead end.

[polyxo]

@polyxo

Interesting all these, we may be out of topic now. A bit.

The reason I'm using some kind of retopo is to be able to have a multi resolution model with UVs.

Hehe... I answered on Rendering only. For this Purpose you wouldn't strictly need a subdividable model.

If you want to keep modeling somewhere else that's an entirely other story.

[polyxo]

Hi All,

I thought I started an own thread for a discussion coming up in Rauls Update thread.

Just as a play of thoughts:

Imagine you start a Sculpting Session with Voxels and Surface Tools.

Now - without previous Retopo and UV-Work - send the Model to the Paint Workspace as Dense Quads.

(3DC would use a much refined implementation of PTEX for this which gave you very nice texture-resolution right away).

You paint Paint Colour and Bump until you realize that something larger has to get changed in the model

(or your Boss tells you that).

No problem - you switch back to Sculpting. 3DCoat displays the Voxel-Object again and you do whatever you like.

(It might be possible to have a viewmode showing the textured Dense Quads and the Voxel-Imput on top of each other.)

When finished with Voxel-Editing the Result was projected onto the already textured Dense Quads.

That should work pretty well as there's a lot of Geometry to work with.

Large topological changes like added Holes might be dealt with by merging painted Dense Quads with a new Dense Quad

Version of the changed portions. All existing painted Ptex information of course had to be kept.

You finish the Texturing of your Model - the object has its final look.

Now you can either export directly and render with Ptex. Or - what any Animator will want to do:

Create a Retopo Mesh of suitable resolution, create UV's and bake Textures from the High Resolution Mesh. Or Retopo but skip the UV's and Bake to Ptex

but onto the Lower Res Version of the Mesh.

Ptex in this case would be used analog to the Vertex-Painting Zbrush uses - with all the Flexibility but the added advantage over the Zbrush-Method that one could

directly utilize its output to render (for Stills at least). The latter option would allow you to use Applinks to create Draft-Renderings inside your Modeler at any time).

Thoughts?

[artman]

Dense quads exports are too heavy to get Ptex assigned to them,Paint room in Ptex mode has a 1-2 million carcass mesh maximum polycount

and at this level performance are far from stellar.

I already do this sometimes for single simple objects.

But it sure would be cool to be able to do this on heavy multi-objects scenes...

[polyxo]

Dense quads exports are too heavy to get Ptex assigned to them,Paint room in Ptex mode has a 1-2 million carcass mesh maximum polycount

and at this level performance are far from stellar.

I already do this sometimes for single simple objects.

Ok, I agree currently painting on high resolution meshes not what 3DCoat is optimized for.

But you know that this is not impossible by principle. Painting on many millions of Polygons is no problem at all in other Painting packages.

Would you see other possible disadvantages - or would the proposal otherwisely make sense to you?

[artman]

Ok, I agree currently painting on high resolution meshes not what 3DCoat is optimized for.

But you know that this is not impossible by principle. Painting on many millions of Polygons is no problem at all in other Painting packages.

Would you see other possible disadvantages - or would the proposal otherwisely make sense to you?

Its not painting on multimillion meshes that is hard its assigning ptex to it.

Once the carcass mesh is assigned you can paint a displaced 10million polys object with 3Dcoat's Ptex mode no problem,

I do it even with my weak computer.

3DCoat is already pushing extremely far Ptex-wise,for example in Mudbox you cannot even assign Ptex to a 200k mesh....that gives you an idea of how far Andrew already pushed it.

Asking for more than 1-2 mil carcass mesh assignment with good performance is probably nearly impossible but if someone could pull it off that would be Andrew for sure.

Probably if he could have done it he would have done it already.

I think this is why Autopo was made ultimately...of course it would be great if it would work perfectly

(with thin surfaces for example) but its probably easier to improve Autopo than assgin Ptex to multimillion meshes.

There is also the "project all" issue...I don't know if you ever tried it on multimillion meshes but it can take sometimes 2 hours...

Edit:Ok,I was able to push as far as 1mil in Mudbox,but thats about the farthest I can get with my cpu.

[Andrew Shpagin]

We will introduce native voxel color painting in vovel/surface mode (in V4), so things will get easier.

[artman]

We will introduce native voxel color painting in vovel/surface mode (in V4), so things will get easier.

Wow! That's some crazy news.

[polyxo]

Its not painting on multimillion meshes that is hard its assigning ptex to it.

I see that this could be a problematic - but wasn't it cool to be able to paint before doing any Retopo-Work?

There is also the "project all" issue...I don't know if you ever tried it on multimillion meshes but it can take sometimes 2 hours...

I just tried a project all on 8 Mio. Took about half a minute in Zbrush.

Ok - this was just a deformed sphere projected onto another sphere...

[polyxo]

Oh - wow indeed!

[michalis]

For this Purpose you wouldn't strictly need a subdividable model.

Indeed. But I need UVs. Uvs need a low poly cage. I tried lot of decimators on 5-10 M meshes. Having a target at ~800k or 1 M. Normally it should work for rendering presentations. Bumps on a 300 k works better. We didn't talk about painting too. Polyxo, you can't UV a 300 k mesh, not even a 100k .

Indeed, I always select to export the appropriate resolution from zbrush and the bumps baked on this. Avoiding displacements which results to similar heavy mesh after all.

Testing the alpha stage of cycles renderer. Displacement modifier is problematic. Normally you keep the 'view' in low poly, "render" on high subdivisions. Cycles reads the render value (as it works in view window, interactive view)

This issue won't be forever but it's the best I have for now.

[michalis]

I was affraid 3dcoat development was commencing it's down phase... Cheers Pilgway

Funny, I was afraid the opposite. It seems I was right.

[chris_solo]

We will introduce native voxel color painting in vovel/surface mode (in V4), so things will get easier.

NICE!!!! very good news!! thanks Andrew!

++Chris

[Spiraloid]

Hi All,

I thought I started an own thread for a discussion coming up in Rauls Update thread.

Just as a play of thoughts:

Imagine you start a Sculpting Session with Voxels and Surface Tools.

Now - without previous Retopo and UV-Work - send the Model to the Paint Workspace as Dense Quads.

(3DC would use a much refined implementation of PTEX for this which gave you very nice texture-resolution right away).

You paint Paint Colour and Bump until you realize that something larger has to get changed in the model

(or your Boss tells you that).

No problem - you switch back to Sculpting. 3DCoat displays the Voxel-Object again and you do whatever you like.

(It might be possible to have a viewmode showing the textured Dense Quads and the Voxel-Imput on top of each other.)

When finished with Voxel-Editing the Result was projected onto the already textured Dense Quads.

That should work pretty well as there's a lot of Geometry to work with.

Large topological changes like added Holes might be dealt with by merging painted Dense Quads with a new Dense Quad

Version of the changed portions. All existing painted Ptex information of course had to be kept.

You finish the Texturing of your Model - the object has its final look.

Now you can either export directly and render with Ptex. Or - what any Animator will want to do:

Create a Retopo Mesh of suitable resolution, create UV's and bake Textures from the High Resolution Mesh. Or Retopo but skip the UV's and Bake to Ptex

but onto the Lower Res Version of the Mesh.

Ptex in this case would be used analog to the Vertex-Painting Zbrush uses - with all the Flexibility but the added advantage over the Zbrush-Method that one could

directly utilize its output to render (for Stills at least). The latter option would allow you to use Applinks to create Draft-Renderings inside your Modeler at any time).

Thoughts?

I like the idea that each room could have a way to update it's changes to the other rooms. the sculpt room seems irrelevant since the vox layers in surface mode is the same thing. but if the UV room and retopo rooms automatically updated all their surface textures via a ptex and texture bake... could be very cool.

I'm picturing:

work on voxels. create a mesh. uv map it. paint on it. voxel sculpt again. the retopo mesh updates it's shape as best it can and you retopo what doesn't work. you enter the paint room and it recalculates your ambient occlusion, normal, displacement and color layers so that even though you have new geometry, new uv's and the new shape, it retains all the pre-existing layers, and paintwork...

I wish this were the case. but painting in the voxel room sounds even cooler.

[polyxo]

work on voxels. create a mesh. uv map it. paint on it. voxel sculpt again. the retopo mesh updates it's shape as best it can and you retopo what doesn't work. you enter the paint room and it recalculates your ambient occlusion, normal, displacement and color layers so that even though you have new geometry, new uv's and the new shape, it retains all the pre-existing layers, and paintwork...

Well I suggested to go further even...

No need UV's and no Retopo until the very End of the entire Creation-Process - so therefore also no requirement to update already begun UV and Retopo work.

To me an updating mechanism sounds more complex and error prone than a Technology which allowed you Game-Guys to do the "down-resing" steps when done with the Design.

I am very curious on direct Voxel Painting too.

[The Candy-floss Kid]

We will introduce native voxel color painting in voxel/surface mode (in V4), so things will get easier.

Oh happy day!! A dream come true.

In the light of all the recent developments within 3D Coat I'd considered it churlish to ask.

It was something Andrew had clearly mentioned way back in the early version 3 but to hear confirmation of painting in voxel mode elevates my spirits enormously as an artist who adores working within the voxel environment.

[Taros]

We will introduce native voxel color painting in vovel/surface mode (in V4), so things will get easier.

Interesting. Really voxel colorpainting or painting on vertex? This is different. What about resolutions? Can I paint thin lines on a low poly voxel object? Or do I paint on the pTex mapped surface model? What happen to my textures, when I switch back to voxels... Much questions and very interesting news.

By the way: About the discussion: There is no application, that allows to paint on very high res textures in realtime and with layers. And I mean real big textures, bigger than 4k incl. 32 Bit colordepth. Mari is the only tool that allows to paint on very high res textures, but not in realtime...

[polyxo]

By the way: About the discussion: There is no application, that allows to paint on very high res textures in realtime and with layers. And I mean real big textures, bigger than 4k incl. 32 Bit colordepth. Mari is the only tool that allows to paint on very high res textures, but not in realtime...

Zbrush actually does.

If you want with 8k and in "HD-Mode" with Models of truly insane Density - Displacement with 32 Bit.

HD however allows you only to edit a small section of the Model at a time though.

This is Zbrush 4 - no idea what they bring in R2.

[Digital777]

In terms of workflow i find it annoying with 3dc sometimes due to the fact there is no true folder dialog so i can't just link new resources but have to manually add them. Maybe it's custom for the brush packs which are all well and good but i would much rather just link to some main folders and download a zip if i want to use other alpha brushes or texture etc.

Also the preset system is not very good and in most other programs my workflow is based around the use of my custom presets which in 3dc i just can't do and have to keep adjusting value each time which again is annoying.

I have mentioned these a few times before and other people have also posted similar things on mantis or the forum but it seems they never get noticed as someone always jumps in saying "i want some better sculpting tools more" etc, so yeah the resource browsing and preset systems really needs updating and i hope these requests won't be ignored.

As for the actual brush/texturing/retopo workflows i quite like what 3dc has already and from what i have seen so far i am sure the liveclay stuff will be amazing also.

[Spiraloid]

We will introduce native voxel color painting in vovel/surface mode (in V4), so things will get easier.

great news!