Jump to content
3DCoat Forums

3D-Coat 3.5 updates thread


Recommended Posts

  • Advanced Member

I tried that but as soon as I go from Surface mode to Voxels I instantly loose all my fine details I had created with Live clay.

I believe that is the case - when you convert back to voxel mode the resolution will only be as high as it was at the current voxel resolution.

If you want the resolution to preserve the details you've added in live clay, I think you have to adjust the voxel resolution by doing a "resample"

Alternatively this is the final step (live clay) before retopo and just never go back to voxel mode.

you can also export the mesh from surface mode - and then import as high res reference mesh.

hope this helps?

Link to comment
Share on other sites

  • Reputable Contributor

Well, right now, if you want to use LiveClay, it only makes sense to do so as a last step before retopology. Otherwise it really buys you nothing. You will have to bake your maps from Surface mode or lose your localized details. I think Andrew has done a good deal of work to improve baking. I was having trouble baking from Surface before his latest work and I think it's gotten better. If there are issues, then I wonder if the mesh didn't break somehow. That's one of the issues they will have to address going forward. It's still early though.

I still hope Andrew is planning on Voxel Subcells. That's where I think the real revolution in the application is at. Some tools work slightly better in Surface mode, but on the whole Voxels are much more forgiving and flexible to work with.

Link to comment
Share on other sites

  • Member

I just finished a job which was mainly about creating digital chocolate and packaging for that.

About 70% was in 3D Coat, while I modeled the rest in Softimage.

So after using 3D Coat for the third time in a commercial job, I have to say there need to be updates to the painting room.

While the functionality is fine the speed is not.

It takes a minute to fill a simple 2k texture with the rectangle tool - this should happen in the blink of an eye.

It took nearly 5 minutes to export 4 2k layers to photoshop.

Any kind of painting larger than a few pixels wide is ultra slow even on simple meshes.

Projection quality needs to be raised a lot aswell.

But the main problem is the speed and filesizes - I had a simple 1.3 million triangle mesh (Microvertex Painting) and a single 2k texture on that,

and the filesize was already at 300 megabytes. There was nothing else on that scene.

Waiting for 3D Coat took more than 2 hours a day while working on this stuff and almost killed me at the deadline - please do something about it.

Oh and please seperate depth, specular and color layers. It is a mess when you have to guess from what layer that specular is coming from, especially

when it takes half a minute to switch a layer on and off.

Link to comment
Share on other sites

Waiting for 3D Coat took more than 2 hours a day while working on this stuff and almost killed me at the deadline - please do something about it.

Oh and please seperate depth, specular and color layers. It is a mess when you have to guess from what layer that specular is coming from, especially

when it takes half a minute to switch a layer on and off.

If you want layers separate, just paint them on separate layers. The problem with that is that you can't paint them all in one brush stroke that way. I do agree about the speed though. my current model takes about 20 seconds just to hide one of the paint layers, about 8 seconds to hide one of the objects. And of course the screen goes all black and says "not responding" during that time, so it looks like it has crashed.

Link to comment
Share on other sites

  • Member

The problem with painting them on seperate layers is that it makes the scenes even slower and bigger (I think).

I thought of a system where you can select your spec, depth and color layers in seperate lists simultaneously.

I think the problem here is that even if you do only paint color an a layer, the depth and specular for that layer is still stored in the file, resulting in 3 maps

per layer, even if only one is needed.

Link to comment
Share on other sites

  • Reputable Contributor

I certainly agree that it's annoying to have to wait for a layer to hide/unhide or to fill a layer. I do hope Andrew will look into that. However, if you're working on a 1.3 million poly model in the paint room, it doesn't surprise me that you would find some lag, to one degree or another. That model would choke Max or Maya, trying to use their built-in painting toolsets. ZBrush and Mudbox may be able to handle dense meshes better when painting on them, but I'm sure you'd notice some performance drop off as well...working with a million and a half polys.

I'm not disagreeing that there needs to be some improvement, but on models less than 100k or so, I've not noticed any real issues with performance. Are you using the latest build (3.5.23C), or a much older one? I can't remember which one it was, but Andrew gave painting performance a kick in the pants several months ago, by multi-threading the Paint Room and introducing some new Intel libraries. I would also like to see better transition times going between the Paint Room and Sculpt Room. It should be near instantaneous, or merge the rooms together.

The Sculpt tools are so few, that they could simply appear in a tools popup.

Link to comment
Share on other sites

Cheers cakeller, I see what you mean, must crank up the space resolution to as high as the fine detailed resolution you created with Live clay, will give that a try.

That's the best short term solution until LC is complete (as in, applied to voxels with sub-cell resolution. Which Raul says is in the works!).

Link to comment
Share on other sites

  • Advanced Member

ZBrush and Mudbox may be able to handle dense meshes better when painting on them, but I'm sure you'd notice some performance drop off as well...working with a million and a half polys.

Fwiw: Actually no (Zbrush) - even in contrast. One can not paint on stuff which is Low Poly - it turns out useless and all blury.

Painting in high quality starts at 4 Mio polys in Zbrush (equals a 2k Map). One can even paint 20 Mio Polys and more (HD-Mode).

Link to comment
Share on other sites

The problem with painting them on seperate layers is that it makes the scenes even slower and bigger (I think).

I thought of a system where you can select your spec, depth and color layers in seperate lists simultaneously.

I think the problem here is that even if you do only paint color an a layer, the depth and specular for that layer is still stored in the file, resulting in 3 maps

per layer, even if only one is needed.

This is an excellent point. I've had lots of issues with this recently.

I think what would be a great solution to this, is, instead of having to globally select which maps to export per layer per UV set to export, a prompt with a pseudo graph layout would be much better, and you could select which layers to export with which maps and with which each UV set. I'm working on a mock-up of this to better illustrate.

Link to comment
Share on other sites

  • Member

I didn't even notice the part about 1.3 million polys. I'm surprised you could even load that, the Paint room isn't meant to handle such high res meshes.

Ah there maybe a missunderstanding here.

I exported a decimated mesh from the voxel room because we did not want to work with displacement here.

The decimated model was 300k triangles (the new decimation algorithm kicked it down from 4 million without loss of quality - great!).

So I went ahead and imported that into PPP, but the cavity painting did not work here - which I needed badly - I guess because it only

reads the cavity values from the normal map, which I did not have.

So I imported it into Microvertex Painting, and the carcass resolution set itself to 1.3 million triangles. Cavity painting

did work, but the whole process was veeery slow, as I already explained above.

Link to comment
Share on other sites

So I imported it into Microvertex Painting, and the carcass resolution set itself to 1.3 million triangles. Cavity painting

did work, but the whole process was veeery slow, as I already explained above.

Yeah I think even if the carcass resolution suggested that amount I probably would have picked something smaller. Actually it's still better to do the retopology or auto-retopo and paint on that, that way you have a nice clean mesh that's low res.

Link to comment
Share on other sites

  • Advanced Member

Yeah I think even if the carcass resolution suggested that amount I probably would have picked something smaller. Actually it's still better to do the retopology or auto-retopo and paint on that, that way you have a nice clean mesh that's low res.

But what if you don't need Low Poly-Meshes in later processes (no Games, no Animation, no Polycount sensitive-Renderer?)

Should one only retopo to be able to paint?

Complex hollow objects are a great example.

It is completely impossible to use Autopo on such objects. Also it may be damn tedious and could take many hours to manually retopo such.

(again - with no actual need for Low Res). In such cases it was great to just send dense quads directly to the Paint-Room with Ptex.

Link to comment
Share on other sites

  • Advanced Member

Well improvements are always better, but I'm curious in what circumstance would you not benefit from having a lower res painted model?

Whenever one needs to create a physical Prototype of what you've done in any Sculpting Program one only needs High-Poly Output.

Low Res meshes are useless in that case.

Say a Product-Designer has created a digital sketch of some Plastic Part in 3DCoat and would like to 3D-print it by giving it to some Service Bureau.

However he might also like to create some Still Renders in order to check some aspects of the Detailing and wants to use the 3DCoat's Paint workspace

for that task. Some Renderers, especially for Still imagery have no problem at all to chew High Polygon Input...

One has to understand that - given the project gets approval - all mesh data (let it be messy high polygon Triangles, Dense Quads, Autoretopo

or carefully made manual Retopo) are of no further use. The project will get entirely redone in Nurbs as industrial level Production methods

without exception prefer Nurbs as Input, some even don't know what meshes are.

So - in non Gaming and non Animation Industries dealing with Retopo may be quite a waste of time as neither High Poly nor Low Poly Meshes are

required downstream. If if the Paint workspace could cope with stuff from the Sculpt Room directly and in equal quality throught use of Ptex I would

probably use it in 95% of the time.

There's special cases like Quad-Meshes converted to Nurbs-compatible Tsplines where Retopo can make also sense for the people outside

of the Digital-Content-Creation Industry though.

  • Like 1
Link to comment
Share on other sites

  • Advanced Member

My apologies, I knew high res meshes were required for rapid prototyping, but it never occurred to me that they would need to be textured.

While it is also possible to 3D-Print in solid colours there's unfortunately no method yet which can replicate real intricate Detail.

So when I spoke of painting I meant additional Render-representation here.

What I mean is that it is questionable why one has to retopo to have liquid interaction in 3DCoat's Paint room when programs

such as Zbrush for instance let us paint on 12Mio Polygons or with tricks even far more.

I on the other hand greatly enjoy that one can paint a Cube with 6 Polys in 3DCoat without having to subdivide.

This is something one can't do with Zbrush...

Link to comment
Share on other sites

I think what zb is doing, and I'm no zb expert, is basically assigning RGB color values to the vertices that are already there, similar to Vertex Paint in LightWave. So therefore there are no huge images taking up memory, but it also has down sides, like only being able to paint on high res and lack of features like Photoshop blending modes. So there are benefits and drawbacks to that system.

Link to comment
Share on other sites

  • Advanced Member

About zb, right phil. But not drawbacks as zb can handle millions of poly, smoothly. Of course there's a drawback in vertex painting, as in 3dcoat MV mode too. This method loves even shaped quads. How to paint a low poly, after retopo (with loops) mesh? Subdivide it as much as you like. Still a problem in some cases. The 4M faces for 2k texture maps isn't exactly precise. Not always.

Link to comment
Share on other sites

  • Reputable Contributor

Try this way.

In retopo room after using autoretopo or the manual method plus making your uv map choose microverts to merge into the paint room. You have used the method before...

In the paintroom under the file menu, choose export and export your model as high polygon quad model. It will capture all your displacement details as real geometry and save the uv map.

You have 3 choices of export sizes to choose from. The sizes are depended upon the carcass resolution you choose in the retopo room. Test it and out and you will see what I mean.

I have exported a 1,450,000 million quad obj model and then reloaded the obj with no subdivision by the PPP method. It took about one and a half gig of ram to hold the model in 3DCoat.

I added a few layers for painting with no problem. Texture maps set a 2048

Linux 64 bit version of 3Dcoat and the latest beta version.

Edit. I tested loaded a 2.6 million quad obj by the PPP method and it loaded fine through PPP but took 2.5 gig of ram to hold the model in the paint room. So you need a good amount of extra ram for your layers and OS system or things will bog down quickly... I was able to paint smoothly...

Link to comment
Share on other sites

  • Advanced Member
It will capture all your displacement details as real geometry and save the uv map.

Not as crisp as it should. A great loss of details. Andrew is working on this, I think so.

Reprojection method in zbrush looks much more detailed, it really is.

In fact, this method (MV) involves smoothed displacement maps, this is how it works. Not as subdivided snapped mesh or similar. Just displacements. I don't need this. I don't like it either.

Link to comment
Share on other sites

  • Advanced Member

Three hours now, testing everything.

It's a great build. Thank you.

Exporting decimated meshes under voxels palette is really great. Such "humble improvements" make my day.

Just minor problems with clone tool in MV mode.

The usual "a little blurred hi poly export" though (MV mode) - displacement maps.

Though LC works fine, I can't do anything serious, tools are needed... tools like sculptris tools, zb tools.

I prefer to work in voxels mode, these simple tools work better for me. But this is only my opinion I guess.

Excellent work Andrew and Sergyi!

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
 Share

×
×
  • Create New...