retopo of complex objects

[Sarichev]

this is really starting to get frustrating,

for some character work i create cloth bases in marvelous designer, my whole reason for starting using 3d coat is that i thought i can import all i need and use it as my main "compositting for the character (meaning organic and hard surface sculpt and paint)

 

i import the cloth bases in S mode, close all holes and convert to a degraded V layer for the Autopo base,

settings for autopo: 15k polygons, 70% detail and 40 for decimate.

 

unless i degrade the model for autopo ALOT or set a low number polys (about 2k) , i get these errors:

 

degrading or going low poly completely missing the point of getting high detail sculpts and then painting low res.

any suggestions/help?

 

thanks a lot,

Sergei

 

 

 

[Tony Nemo]

Why convert to voxels? Try this: in S mode, Decimate to 70% and proceed with Autopo. Try with no adjustments initially.

[Sarichev]

Tried it, 3D coat crashes or just gets stuck not doing anything 

[Carlosan]

Hi

Are you using the last version 4.5.40 ?

[Sarichev]

It's 4.5.19 edu 64bit

[Carlosan]

The new stable version is v4.5.40 with several bug fixes.

[AbnRanger]

I always try to run Auto-Retopo on a more "dumbed-down," less-complex/detailed version of the model. Once you have successfully used Auto-Retopo, then you can discard the dummy version. Reason is, it gives the algorithm less guess work and fewer calculations to perform, fretting over details that normal/displacement maps are going to provide later in the pipeline.

It can work pretty well in the right situation. You can try fewer guides than in this video, and it should do alright

 

[Sarichev]

i upgraded to the newer version, still this barely works. 

it either gives a bad retopo or this error:

[AbnRanger]

Can you send this file to Andrew through the HELP menu? Then e-mail support@3dcoat.com to let him know you just sent it to their server and the details. Sometimes, 3D Coat has a real problem with a given model. I had that occur on the 3D Scan mesh, even after I cleaned it up. I made a lower res duplicate and it had no problem with it. 

[Carlosan]

Your model is very small. Try scaling x3.

Anyway, do you reported the crash to support@3dcoat.com ?

Please attach the model or share a download link in the email.

Ty.

[Sarichev]

i attached the link to the model.

https://www.dropbox.com/s/7th1zi8teljy1c1/tests12.obj?dl=0

 

thanks ill send these to the support/help

 

 

[AbnRanger]

It is a problematic mesh for Auto-Retopo to handle, because it has so many self-intersections, and a few rough spots/blobs up top where the helmet would sit. However, I think you were on the right track, originally. Going the retopo/auto-retopo route is just totally unnecessary for what you are trying to do. It's better to just vertex paint on the high-poly. I subdivided the mesh you provided, and then tried to apply a Smart material and noticed a bug.

You were right, it is not applying it correctly. I couldn't get it to apply the glossiness, even though the glossiness level was up to 100% and the channel was enabled. Please send another e-mail to Andrew and let him know this is the real source of your problem and frustration. I'll try to get in touch with him about this tonight or tomorrow, about this subject, cause it needs to get fixed ASAP.

[AbnRanger]

Just did a little more testing on a new scene where I imported the same model and subdivided (+ smooth all) it > then applied a Smart Material to it. Worked correctly, but got borked when Auto-save kicked in while having the preview window open. I'll let Andrew know about it.

[Sarichev]

Thx a lot of trying,

if I do go straight to paint room I can't export  obj with t4xtures right? Because it doesn't paint on any  UVs 

if I take the model to zbrush zremesh to lower count and dynamesh I get the retopo fast just need to find a way to put it all back. Very messy

[Sarichev]

When I export these objects from marvellous I can get them with UVs and can paint on them in paint room with no issue.

 

what I wanted to do Is to tweak the geometry in sculpt room more to make it fit other objects and then paint. That's why I import into sculpt first and then trying to re topo

[AbnRanger]

1 hour ago, Sarichev said:

When I export these objects from marvellous I can get them with UVs and can paint on them in paint room with no issue.

 

what I wanted to do Is to tweak the geometry in sculpt room more to make it fit other objects and then paint. That's why I import into sculpt first and then trying to re topo

If you get a clean, low-poly quad mesh from MD with UV's you can follow this series of steps:

1) Import Mesh into Retopo Room first

2) In Sculpt room, under GEOMETRY menu > Retopo Mesh to Sculpt Mesh (new in build 4.5.40)

You now have your low poly mesh in the Retop room and High Poly in Sculpt room. As you make large-scale edits to your Voxel/Surface sculpt, you can use the CONFORM RETOPO MESH option. With smaller-scale edits, like brush strokes and such, you will periodically want to step into the Retopo Room and either click SNAP for the entire mesh or do it in a localized fashion, using the BRUSH tool (or SELECT tool > make specific selection and then SNAP in SELECTED section of the Tool Panel)

BTW, I thought you were intending to paint on the high poly mesh and just render the objects in the Render Room > composite in Photoshop for Concept/Illustration work?

[AbnRanger]

Another tip might be to make sure your exported meshes from MD are free of self-intersections. It seems to do a good job with Cloth simulations, but maybe the THICKNESS and SELF-INTERSECTION settings need to be modified to help prevent these issues. They are problematic in no matter what application you use.

If it were not for the number of self-intersecting parts of the mesh, and the really thin area where the helmet would rest on top of, Voxels (or Surface mode) could handle this job very well. Those were the two issues I had with it, when testing. Auto-Retopo needs some degree of thickness, like Voxels do. Giving it a mesh where you have some paper thin sections, is inviting problems. It may be frustrating to find a good workflow, but once you do, it should go a lot smoother for future projects.

You're still learning how things work, and it can be aggravating trying to get things done during an exploratory process. But, if you give 3D Coat a decent piece of geometry to work with, it can do an amazing job, whether it's sculpting, painting or Retopo work.

[Sarichev]

thanks a lot,

yea thats what i was doing but then scenes get bigger and i had to export to MAX or keyshot. it kind of made me do extra double or triple work. i invest the time in sculpting and painting the model in 3dcoat but then basically if i need it exported to a bigger scene im stuck and have to redo it piece by piece all over again.

i think thats why i still need the UVs after all,

now i tried to simplify the model in zbrush , dynameshed it, the model is much less complex and mesh is closed but still 3dcoat cant retopo it. software either fails to retopo or crashes.

here is a link to the model, its very simple now but still cant make it work:

https://www.dropbox.com/s/vqszifwv34yc0qu/zbrush tests16.OBJ?dl=0

 

a workaround im trying to test now basically goes through importing all to the sculpt room, sculpt whatever else is needed and combine, i then export back to zbrush for retopo and import back to 3dcoat to paint room (but now i have the UVs made in zbrush).

it kind of works but i just dont get why it has to be so complicated work around.

 

why i go though all this mess? i really like 3dcoats sculpt freedom for both organic and hard surface i save a tone of time. its better than MAX/maya polygon clicking and better than zbrush for edge control. Also, i like the paint options for texturing, over BPR max/keyshot textures by far. i just need to find that friggin bridge between those to make it work all in 1.

sorry that its kinda irritating but atleast i got you interested

 

 

 

Edited June 4, 2016 by Sarichev

[AbnRanger]

3 minutes ago, Sarichev said:

thanks a lot,

yea thats what i was doing but then scenes get bigger and i had to export to MAX or keyshot. it kind of made me do extra double or triple work. i invest the time in sculpting and painting the model in 3dcoat but then basically if i need it exported to a bigger scene im stuck and have to redo it piece by piece all over again.

i think thats why i still need the UVs after all,

now i tried to simplify the model in zbrush , dynameshed it, the model is much less complex and mesh is closed but still 3dcoat cant retopo it. software either fails to retopo or crashes.

here is a link to the model, its very simple now but still cant make it work:

https://www.dropbox.com/s/vqszifwv34yc0qu/zbrush tests16.OBJ?dl=0

 

a workaround im trying to test now basically goes through importing all to the sculpt room, sculpt whatever else is needed and combine, i then export back to zbrush for retopo and import back to 3dcoat to paint room (but now i have the UVs made in zbrush).

it kind of works but i just dont get why it has to be so complicated work around.

 

sorry that its kinda irritating but atleast i got you interested

 

 

 

Actually, you could handle a pretty big scene in 3D Coat, depending on your hardware. I noticed you have a GTX 760. It is basically a first generation Kepler (architecture) Nvidia card. I bought a 4GB Gtx 670 a few years ago and after  a month, I ended up selling it on eBay. The reason was, it had terrible performance when wireframe was turned on in the Sculpt room. In fact, it was much worse than the GTX 470 it was replacing. I ended up going back to the Fermi architecture, which handled dense 3D meshes much better and had much better CUDA performance. One major reason was because NVidia tends to get stingy with the Memory bus sizes. The GTX 580 had a 384bit memory bus, while the GTX 6xx-770 line went backwards to a 256bit memory bus. Larger is better.

They sacrificed performance for the sake of energy efficiency. I also think it is because the larger memory bus likely uses more gold or copper = more expensive to make. This is why, no matter how much fanfare a new NVidia line may come with, I ain't buying another card of theirs with a 256bit memory bus. They may have generated enough software hacks to make games perform better despite the small bus, but you can't fake it when working in Conent creation, with Millions of Polygons and large texture map sizes.

I said all of that to say this....it might be time to try and upgrade to a 980ti (which should drop dramatically in price, now that the 1070 beats it for a lot less) or try the 1070 and see if it handles large scenes in 3D Coat fairly well. Actually, it will handle it just fine, but when wireframe is turned on, it might be a bit sluggish because of the 256bit memory bus limitation.

Another hardware consideration is RAM. You really need 16GB to 32GB to handle relatively large scenes in 3D Coat, and it's pushing it with 16GB. But remember, you can always cache layers you are not actively working on. When 3D Coat temporarily swaps the original version with a decimated proxy, it will look nearly identical to the original, and you can work on it, if needed. This can reduce your memory load substantially. Hiding objects/layers you aren't working on, will also reduce the load on your graphics card, so you will notice a sizeable improvement in performance if you both, cache non-active layers, and hide them.

[Carlosan]

I contacted Andrew about this issue, he wrote:

Some meshes may not be autoretopo-ed.

Knowing limitations may help to prepare shape.

Multiple inner holes, too complex topo-structure, unclose-ness of the mash may lead to failure of algorithm.

Andrew

[Sarichev]

Yeah I'm waiting for a Titan x 12gig to arrive, I got 32gig ram but I'm not planning to go to 64gig just yet want to see how much the Gpu will handle

[AbnRanger]

32GB should be fine. That's what I am using and it is more than sufficient for almost everything I need to do. So, let me try to understand why you need to retopologize rather than just use vertex paint in 3D Coat. You have a larger scene and think you will need your objects to be lower poly than what you have, currently? Again, you can effectively decimate the objects/layers in the scene, to really low levels by caching them. Uncaching them will restore the object to it's original resolution state and translate any changes made to the proxy, to the original.

In this video, he appears to be able to work with a pretty heavy scene:

 

[Sarichev]

I don't mind the size of the objects as long the render works, I had a scene with like 15gig, with character objects of 0.5gig or so and its fine. If I do vertex paint I ca not get that painted info out of 3D coat right? I use the 3d coat  renderer sometimes but usually it's either I'm more used to external renderers or I need them because they have more options so exporting becomes important for me. I'm going to run some heavy tests in 3D coat once I get my 2nd Gpu to see what I can get away with without leaving 3D coat.

 

so, my "nice to have" from 3D coat would be to do the full sculpt (including cloth which is MD) & texturing in it, the higher poly the better.

for character design iterations  I stay in 3D coat and just render inside as well.

But once that phase is done and I build a scene I export everything  to keyshot or max. It includes also props/ground/bg assets all that stuff I model in3dcoat and use procedural textures in keyshot/max. For the peripheral , procedural textures are fine but for hero asset or character it's either I spend a lot of time painting or I get a good textured base that saves me tons of time once the scene needs to be changed. That's why I'm breaking my head here.

 

Hope I made some sense and didn't babble too much.

thanks a lot man

 

 

Edited June 4, 2016 by Sarichev

[AbnRanger]

Not every 3rd party renderer supports vertex paint, but as you can see in the video, here....it should work in 3ds Max, VRay or Mental Ray. iRay might support vertex color, but I'm not sure. When you get your new graphics card, I would honestly explore using iRay from here on out. It's pretty awesome. Yes, it has a few limitations such as not working with volumetrics, but it is otherwise pretty darn cool. Keyshot is all CPU based, IIRC.

 

 

[AbnRanger]

I was going to try and show a recording of how one could simplify the model you provided and voxelize it, but to be honest, the mesh Marvelous Designer produced is just too problematic. I was having to use all the tricks I could think of, to get around all the self-intersections and voids, and had to resort to using the Fill Tool to try and clean a lot of that up, but then the video got over 30min, so I scrapped it. Sort of defeats the purpose of making things easier/faster to use for Concept Art/Illustration work.

If it were my project, I would run the cloth sim again in MD, but try to increase the cloth thickness parameter, and see if there isn't some control over self intersections. The other thing to keep in mind is, for concept art, you probably don't need uber-dense texture resolution, so you could just import the mesh as a Surface mode object > using the SUBDIVIDE tool (top of the Tool Panel) to apply one or two iterations of Subdivision to the mesh > paint in Paint room > export (as FBX) from Sculpt room. Vertex color will be stored with the model. Many of the other elements in the scene can be Auto-Retopologized and UV'ed, to keep your scene as light as possible.