V4.1 BETA (experimental 4.1.17D)

[BeatKitano]

So far the safest for me has been: save another file before doing anything to "fix it".

 

Then close holes and reconstruct if there's still non manifold geometry. Cleanup memory I don't trust, most of the times it creates tears in other places.

Tears you don't see right away and bite you in the butt after a while (mostly unfixable because the mesh is so fragmented you fix a part your destroy another you don't see or because brushing over teared parts create mesh explosions).

Since 4.0.6 it's been my process and I've not used it often (pretty stable on my side but as always it depends on your workflow).

[digman]

Hey Beat, how does the new pinch work? I still waiting for 4.0.06 to be uploaded for the Linux...

 

Yeah, I always save my files before fixing, a standard in 3DCoat and for that matter any program.

 

All those fixes under the voxel menu still need a looking over by Raul when he gets back to get them working without bugs.

 

Close holes and reconstruct does seem to the be the safer route atm...

[BeatKitano]

@Digman

 

It's very nice, before you needed large radius to really affect the pinching process. Now you can use Artman's preset to get a quick pinch with the desired size (avoiding to eat everything in the large radius you had to use before).

In fact, maybe I'm just getting used to it but I said I wasn't convinced by the sharp earlier, I made a custom one based on Artman's settings without steady stroke, higher smoothing and lower detail and works just as good as zbrush one.

In fact I like it more because Dam_Standard in zbrush is stressing the mesh topology but with removestretching it makes it harder to make a mess of the surface.

There's still the interpolation issue with his other Scifi serie, but maybe it's just about settings if I use his curve and the same kind of setting as my sharp it may fee ok (gotta test, but I don't have much time those last few days).

[ajz3d]

Okay, I'll pack it up and send it to Andrew. Thanks.

 

By the way, I have noticed that after sculpting for a while with Clay (or any preset that utilises it) , brushing with this particular tool stops affecting the mesh and I have to reload the scene to make the Clay work again. I don't know if it matters, but I jump a lot between Artman's: Clay, Fast Pinch, TSmooth and Tweak presets and the default Clay.

 

[BeatKitano]

I got an issue like that, it may be a bug in 3dcoat but I'm not sure: the pen pressure is not detected anymore.

It means 3dcoat always see 0 pressure (even using mouse doesn't work). 

In that case, what I do is reactivate detection by either clicking on the ui panels ion 3dcoat, or clicking in another window and clicking again in 3dcoat.

In my case I guess it's more a driver issue (wacom driver not knowing in which app I'm in) but I thought it would be good to tell it here in case someone may encounter the same issue (which may be your case ?).

 

Also if you're running the cuda version : try simple I got weird thing like that (this includes tears/holes/mesh explosions btw!) in the past on cuda, never used it again and to be honest last time I checked the difference was not really noticeable(but I'm using surface mode mostly !).

[ajz3d]

Could be, though I'm fairly sure other brushes do detect the pressure. I'll double check next time it happens, just to be sure.

[BeatKitano]

Could be, though I'm fairly sure other brushes do detect the pressure. I'll double check next time it happens, just to be sure.

Sounds like my issue, before I discovered it was a focus issue I switched to other tools (some would work) to reactivate the brush I wanted to use .

 

But complete failing to use your brushes even after doing that is an issue I've experienced in the past after a massive mesh corruption. Often because of tears in my mesh. Try smoothing out the area you're trying to brush to see if you spot some. If that's the case I strongly suggest leaving the cuda world and fixing your mesh in simple version.

 

Also tell me if I'm wrong: If you get tears/holes/explosions it's on a mesh you switched surface<>voxel twice at least, right ?

[ajz3d]

(...)

But complete failing to use your brushes even after doing that is an issue I've experienced in the past after a massive mesh corruption. Often because of tears in my mesh. Try smoothing out the area you're trying to brush to see if you spot some. If that's the case I strongly suggest leaving the cuda world and fixing your mesh in simple version.

 

Also tell me if I'm wrong: If you get tears/holes/explosions it's on a mesh you switched surface<>voxel twice at least, right ?

You're right. I did convert this mesh a couple of times from voxels to surface and back again. One particular explosion I have described in one of my posts above, though I was able to fix it eventually by slightly lowering the resolution of the voxel volume and only then converting it to surface. After this, Clean Surface didn't explode the model any more and so far I get no more tris popping right into my face.

 

I've fixed the problem with those two holes using CleanClay set to Reconstruct before revealing them (Close Holes applied after they showed up seemed to freeze the program forever) and, by the way, LiveClay brush made them appear too, so it wasn't limited exclusively to CreaseClay. I think it's like you said: the corruption happens somewhere during conversions. Some of LC brushes just happen to reveal it.

[digman]

Always run fill voids and close invisible hulls under the vox tree tab after switching to voxel mode from surface mode. Do it each time you switch... It helps a lot in getting rid of surface mode corruption that is revelled at some point when you switch back to surface mode. That is my experience so far...

 

I have had some mesh explosions when switching back to surface mode but it seems those two tools take care of them in most cases. That is my experience so far...

[BeatKitano]

You're right. I did convert this mesh a couple of times from voxels to surface and back again. One particular explosion I have described in one of my posts above, though I was able to fix it eventually by slightly lowering the resolution of the voxel volume and only then converting it to surface. After this, Clean Surface didn't explode the model any more and so far I get no more tris popping right into my face.

 

I've fixed the problem with those two holes using CleanClay set to Reconstruct before revealing them (Close Holes applied after they showed up seemed to freeze the program forever) and, by the way, LiveClay brush made them appear too, so it wasn't limited exclusively to CreaseClay. I think it's like you said: the corruption happens somewhere during conversions. Some of LC brushes just happen to reveal it.

 

My guess: voxels artefacts are converted to non manifold geometry when converted to surface mode. That's why it's random the vox artefacts are completely random and sometimes you don't see them right away, when converted to surface they disappear (it's an enveloppe) and get converted to non manifold geo, triggering all kind of nasty ranging from explosions/holes/tears.

 

Start in voxel finish in surface or start in surface, convert to voxel and back to surface but only once, after that it's risky business.

[Ellie123]

Just tried the split joints tool for 3d printing and it seems to work very nicely. Though I am wondering about a few items and if anyone else was testing it.

my workflow is usually - to go from doing all booleans in V mode and then at the end (after I have made hull) go into S mode for small details and additional tweaks.

SO now I'm liking this new joints tool in S mode but I'm wondering has anyone used it for 3d printed stuff and where in workflow they are using it?

Would not want to go back into V mode after making a joint I'm worried about inaccuracies with shrinkage on the joints.

Anyway i'll post back if I run a real test on this tool with a real print to confirm- but I am a bit nervous since my prints tend to be on the bigger side and don't like to make expensive mistakes.

So anyone tested this yet for real ?

[PolyHertz]

I just tried the 'Paint With Splines' tool, and I cant see the line width? And when the stroke is applied its a solid line (no smoothing/aa) like it was done by the pixel brush, that is when it applies strokes at all. Anyone else having the same issues with it?

 

Also I typically just paint color textures in 3DC, but was trying depth painting and for some reason the layer opacity doesn't affect it at all? Is that normal?

Edited July 25, 2013 by PolyHertz

[PolyHertz]

Another issue; The paint rooms Transform/Copy tool projects the paint with an offset if the transform box is skewed instead of rectangular. If the select box is skewed to the point of being concave the resulting projection will appear somewhat corrupt.

[digman]

I normally do not ask this but can the Linux and Mac experimental versions follow a little closer behind the windows version. There are improvements that get added, that we as well would like to test out.

 

Now, I know the windows version needs to be always first to find major problems and fix them before the linux and mac versions are compiled as I think you would have to pay for each compiling to the Linux and Mac programmer.

 

It just seems lately we are falling further and further behind... Thanks in advance for your consideration...

 

Oh by the way, also thanks for getting some time to work over Autoretopology as stated in your Twitter postings. I know it will improve alot.

[SilverCity]

I normally do not ask this but can the Linux and Mac experimental versions follow a little closer behind the windows version. There are improvements that get added, that we as well would like to test out.

+1

[mercy]

"Andrew Shpagin ‏@AndrewShpagin 19h

4) Very accurate retopology in case of symmetry enabled."

We need 100% quad  mesh, because if your Autoretopo routine creates a triangle in the fingers or in a place remote from the symmetry line, we will have trouble making that mesh 100% quad again. There is a separate category for quad products on Turbosquid for example.

Quad is easiest to sculpt on. 

Thought about this a bit and maybe if your autoretopo is very good - better than Zbrush R6 - we can fix those triangles by hand. Detaching triangle containing mesh part and re-modelling it anew then attaching it back.

Edited July 26, 2013 by mercy

[BeatKitano]

"Andrew Shpagin ‏@AndrewShpagin 19h

4) Very accurate retopology in case of symmetry enabled."

Quad is easiest to sculpt on. 

 

That's where this news leaves me puzzled. I don't want to say Andrew is doing this for nothing (there'll always be use case where this comes very handy), but for now 3dc don't make use of subdiv levels (main reason why quads are prefered along with animation).

So if 3dc is getting in a fight match with zbrush about autoretopo it's ok for me, but contrary to zbrush I don't see an immediate gain from it.

I don't expect it to output perfect topology (zbrush doesn't and I'm not sure anyone at the time being is able to pull that off).

In zbrush zremesher was necessary because dynamesh can only get you so far before slowing you down a lot and you need a proper topology+levels to get a good experience. In 3dc we're free au topological consideration (if we want to) we don't rely on subdivs.

 

If it's an external output for animation: I wouldn't take most of zremesher output for that and I expect the same for 3dc.

So... basicaly: why ?

[AbnRanger]

"Andrew Shpagin ‏@AndrewShpagin 19h

4) Very accurate retopology in case of symmetry enabled."

We need 100% quad  mesh, because if your Autoretopo routine creates a triangle in the fingers or in a place remote from the symmetry line, we will have trouble making that mesh 100% quad again. There is a separate category for quad products on Turbosquid for example.

Quad is easiest to sculpt on. 

Thought about this a bit and maybe if your autoretopo is very good - better than Zbrush R6 - we can fix those triangles by hand. Detaching triangle containing mesh part and re-modelling it anew then attaching it back.

I'd rather have one triangle to deal with, than a nasty spiral loop.

[artman]

If it's an external output for animation: I wouldn't take most of zremesher output for that and I expect the same for 3dc.

So... basicaly: why ?

-To get a decent mesh to test shaders in game engine

 

-To get a mesh to paint for render purpose using displaced Ptex or 1 million polys PPP.

 

-For landscape and static objects...

 

ect...

 

It can also save tons of retopo time work  as it is not true that its always better to start from scratch...correcting the edge flow of zremeshed objects is less long than retopologizing from scratch even if its something people dont believe yet...its the truth.

[digman]

A couple of reasons for better auto retopology:

 

Makes selecting your uv seams much easier, spirals do not help at all in this process and you can not create a very good uv map with them.

Edge loops following the curvature of the model better for static objects in a scene. If I can get a good enough mesh for some static objects then more time can be spent else where.

 

I have never looked at auto retopo as a end all for manual work but it sure has it advantages in some cases so if Andrew can improve the routine, more power to him...

 

Manual retopo is here to stay and I would I think be best for objects close to the camera in your scene and animated objects but auto retopo is here to stay too so it needed a little TLC...

[BeatKitano]

-To get a decent mesh to test shaders in game engine

 

-To get a mesh to paint for render purpose using displaced Ptex or 1 million polys PPP.

 

-For landscape and static objects...

 

ect...

 

It can also save tons of retopo time work  as it is not true that its always better to start from scratch...correcting the edge flow of zremeshed objects is less long than retopologizing from scratch even if its something people dont believe yet...its the truth.

1)You don't need perfect loops for that, crank the resolution high enough and get the same result (preview not production...)

2)Could be a reason, but then again it would mean the algorithm needs to be very high quality for that to work out the box. And I mean "could" because at render stage every mesh is triangulated by the renderer, in the end the overal distribution is more important than loops (unless you plan to animate).

3) static mesh don't need perfect loops. In fact since you don't animate you don't even needs loops at all, only matching shapes. It's true with both rendering and game engines (tesselation).

 

1 reason out of 3.

 

As for starting from a zremeshed, well you can start by creating your retopo from scratch from singular points and trust me, you'll make your topo in VERY little time (a few hundreds to few thousands poly cage sudvided as needed)... in fact most of the time it's quicker than correcting errors from an autoporetopoed mesh and the result will flow much better.

But then I'm fighting the wrong fight for sure as it's a novelty everyone will want to try it before deciding for themselve.

 

Note: I'm not against it, it's just I don't see the emergency here and it looks like one of those "rushed" moment where Andrew wants to have the same tool as the competition while he could simply ignore it and go with his own paradigm (which he started and zbrush has been copying with dynamesh for instance), there's tons more thing that need attention (Polyhertz breaking my heart with all his good paint room requests for instance)

[AbnRanger]

1)You don't need perfect loops for that, crank the resolution high enough and get the same result (preview not production...)

2)Could be a reason, but then again it would mean the algorithm needs to be very high quality for that to work out the box. And I mean "could" because at render stage every mesh is triangulated by the renderer, in the end the overal distribution is more important than loops (unless you plan to animate).

3) static mesh don't need perfect loops. In fact since you don't animate you don't even needs loops at all, only matching shapes. It's true with both rendering and game engines (tesselation).

 

1 reason out of 3.

 

As for starting from a zremeshed, well you can start by creating your retopo from scratch from singular points and trust me, you'll make your topo in VERY little time (a few hundreds to few thousands poly cage sudvided as needed)... in fact most of the time it's quicker than correcting errors from an autoporetopoed mesh and the result will flow much better.

But then I'm fighting the wrong fight for sure as it's a novelty everyone will want to try it before deciding for themselve.

 

Note: I'm not against it, it's just I don't see the emergency here and it looks like one of those "rushed" moment where Andrew wants to have the same tool as the competition while he could simply ignore it and go with his own paradigm (which he started and zbrush has been copying with dynamesh for instance), there's tons more thing that need attention (Polyhertz breaking my heart with all his paint room requests for instance)

He said he's going to work on the Paint Room soon, but I personally would rather he answer what Pixologic did, sooner rather than later. They stirred up a lot of buzz over that...notice it was done soon after V4 was released. So, no doubt they were trying to steal some of 3D Coat's thunder/momentum. I'm glad he's returning the favor.   And it's not simply doing it to answer back...but they showed that the algorithm could be vastly improved.

 

I think using Retopo Presets, stored in the Models Pallet, is the best way to go, anyhow. It's quick, and it's accurate. But it would be nice to know that you can rely on Auto-Retopo to do the job, when and if you need it. I would use it more often, if it didn't have all those gotcha's. It can come in handy for quick prototype work or using 3D Coat to model text and graphics for Motion Graphic elements, too. I know I would often have to clean up text in Max, before I could use it in rendering. The long skinny faces and Ngons that are naturally produced when using the Text tool and extruding or beveling them, make an otherwise easy task more laborious than it might appear.

 

Doing this in Voxels instead and getting a clean Auto Retopo result could be a boon. It also would make Hard-surface modeling MUCH more feasible to do in Voxels.

[BeatKitano]

The thing AbnRanger about "getting back" is that Zbrush is the leader, if they do zremesher, even if zremesher is a copycat of autopo it's still pixologic marketing which will be in all minds... 

Improving I'm all for, but getting into it just to prove it can be better in 3dcoat won't matter ONE BIT (I'd love to have multi board poll about knowledge of autopo vs zremesher).

 

Now creating another "handle" on things like Andrew did with freeform sculpting with the combo vox/lc is another path one not clouded by marketing tricks of the giant of the moment.

 

About the use cases: motion graphics could indeed benefit that, I'm much more convinced here

I don't understand your point about vox hardsurface though, you mean sculpt rough>retopo>resculpt ? You would lose a lot in the conversion process, even with high resolutions since you also need to make a very dense mesh with your retopo mesh.

 

Or did I get it wrong ?

 

You all somehow convinced me there's still benefit in it, but I really think this should not be a rushed feature. The thing Andrew did with 4.0 for instance hurt sales I'm sure of it, for instance if he had released 4 with the latest beta finish it would've made a much bigger splash. Priorities

[digman]

Proper edges are important even on a static mesh.  When you have polygons crossing the model shapes at weird angles, your normal maps do not come out as well. Also you can not cut the uv seams in a logical fashion to help avoid artifacts which is even more important for displacement maps.

 

For a normal map, I do not want to boost my subdivision levels. defeats the purpose of a normal map.

 

Sure, I have / had to wait for things (Paint room) like the rest of us but maybe the Zremesher inspired Andrew to get better results in his autoretopo, not really a bad thing for a developer.

[BeatKitano]

Proper edges are important even on a static mesh.  When you have polygons crossing the model shapes at weird angles, your normal maps do not come out as well. Also you can not cut the uv seams in a logical fashion to help avoid artifacts which is even more important for displacement maps.

 

For a normal map, I do not want to boost my subdivision levels. defeats the purpose of a normal map.

 

That's where I disagree, completely.

What matter if you're not doing displacement and not animating is silhouette. It doesn't matter how the underlying structure is done as long as the shape retains its quality under "every" angle. And this not necessarily goes trough edge loops. IT's actually more about tris orientation (those polygons crossing the model shapes at weird angles) If you go quad only you'll get those at places, if you go full tris you can control that.

 

As for RT work: you don't ask for a quad mesh for optimisation purpose. You work with tris and quads, and even only tris if you want to get the best shape possible with the less amount of polies. Beside you don't even need to care about quads nowaday as tesselation is about to get the norm and it works for most engine with tris (only opensubdiv like tesselation use quad based subdiv). Again all my arguments are for NON animated objects. And if you want to use autoretopo (this goes for 3dc and zb btw) for animation purpose you're doing it wrong, you WON'T get really loop friendly meshes, only "meh" ones.

[digman]

Having to cut uv seams across weird angle polygons over a model's shape does make a difference. I had some uv seams showing up in the normal maps because of this... Better edge flow does help to reduce seam problems so It does matter.

 

Also I think it's make a big difference in your renderer. A low polygon mesh with bad edge flow, (non-planar polygons)  does have more artifacts under GI lighting. I know, I have seen it many times...

 

You said what if your are not doing displacements. Displacements are a part of some artists 3D workflow so it has to be taken into account in 3DC's overall autoretopo routine. 

 

Yes, I have been talking also about static meshes only.

[BeatKitano]

"I had some uv seams showing up in the normal maps because of this... Better edge flow does help to reduce seam problems so It does matter."

 

True, but in this case it's not about the tris seam only, it's about your uv and normal orientation. If you don't plan right you'll get vector issues (orientation conflict mainly). Now I agree that proper so called "edge flow"  (because for statics an edgeflow is always broken with tesselation, i would prefer to call it poly distribution) it can be reduced but you can't say "perfect edge flow make seams almost invisible", other factors are here.

 

I agree about GI, but if you're using GI it means either you're using RT/baked gi in this case it doesn't really matters (previous point), or render GI in that case you do a proper edge flow from the get go or you're doing sloppy/lazy (I'm fine with it under certain production condition !).

Same can be said about displacement.

[digman]

@ Beat, Good discussion here as always even though we tend to disagree at times but hey life would darn mundane if we agreeded on everything...

 

Yes, edge flow or proper polygon distribution is not the complete answer because many factors play into getting good maps and then the rendering part..

[AbnRanger]

That's where I disagree, completely.

What matter if you're not doing displacement and not animating is silhouette. It doesn't matter how the underlying structure is done as long as the shape retains its quality under "every" angle. And this not necessarily goes trough edge loops. IT's actually more about tris orientation (those polygons crossing the model shapes at weird angles) If you go quad only you'll get those at places, if you go full tris you can control that.

 

As for RT work: you don't ask for a quad mesh for optimisation purpose. You work with tris and quads, and even only tris if you want to get the best shape possible with the less amount of polies. Beside you don't even need to care about quads nowaday as tesselation is about to get the norm and it works for most engine with tris (only opensubdiv like tesselation use quad based subdiv). Again all my arguments are for NON animated objects. And if you want to use autoretopo (this goes for 3dc and zb btw) for animation purpose you're doing it wrong, you WON'T get really loop friendly meshes, only "meh" ones.

I think you're wrong in this instance. It's true of the current state of Auto-Retopo, but from everything I saw in the Z-Remesher videos was that it does a MUCH better job and gives a cleaner result, regardless of whether it will be animated or not. And what I meant about Hard surface modeling, is that 3D Coat's Voxel modeling tools provide a very good alternative to modeling in a traditional modeling app....but the downside has always been that you end up having to retopologize the model...often manually. That deters a lot of folks, who don't see the benefit. If Auto Retopo is far more accurate and usable, then it eliminates that issue.

 

In fact, it already does a pretty good job on hard surface objects...but with spiral loops and other issues removed, I can see it being an invaluable toolset for this kind of work. Yes, modeling in a traditional 3D App might make more sense in many cases, but I always hated having to piddle with verts, edges and polys...and have to weld overlapping verts, and just the normal clean up involved, throughout the whole modeling process. So, if you model in the Voxel Room, you eliminate a lot of that fuss, and can just create. An accurate, clean Auto-Retopo routine would allow me to bypass virtually all of that piddling and cleanup involved. I think a lot of studios would stand up and take notice...because Art Directors and Studio owners would like to get rid of the piddling and fussing with geometry too.

 

The way I look at it, is this....Pixologic made a name for themselves because they introduced a strange but effective way to do things that just wasn't possible before. I think Andrew can make a name for himself by continuing to make 3D Coat a very serious modeling alternative. No one ever thought of Voxels as a modeling platform, before now. Andrew is pioneering through new territory in this regard. It makes sense to ensure that Auto-Retopo lives up to it's promise...cause beforehand, it had not.

 

One of the reasons I made sure to include a lot of modeling material in the V4 promo was because almost nobody viewed 3D Coat as a capable modeler.

[BeatKitano]

@Digman:

 

:)My point to summarize (because this has been dragging for too long and we're all using the software in different "sectors" thus obviously can't agree on everything): if Andrew is reworking autopo for animation mesh: it will be "meh" at best. 

If he's working on it for sculpting: I don't get it why reinstate quad based/subdiv paradigm while we have freeform. If there was a subdiv room I'd would be all for it, just like AbnRanger, it would open a lot of doors with all the conversion processes in 3dc !

If he's working on it just to do like zbrush: I don't get it, seems pointless to me.

If he's working for other purpose (output mesh for statics more accurate) why not. But I don't see an emergency. That's all.