Ideas about a new 3DC

[michalis]

Please be careful, this is not a wish list!

Its about a new 3DC mostly.

1. Trying to capture details from a tri model using a quadrangulated cage is very difficult and a great challenge.

The problem with snapping exists here, the problem with baking too. Some times it looks like normals are inverted when baking, (but they aren't). Especially near openings, around ears or anywhere snapping is an issue.

2. The best way I know for subdividing is the "catmull-clark" method.

3. There must be a "room" after having a heavy subdivided mesh where traditional surface sculpture is possible (a small zbrush). It already exists in 3DC (sculpt room) but here some great tools are missing and most important its not a multi-res room. The great advantage of ZB is here, tools always close to what artists want. Blender team are already working on this for example. Capable to handle millions of poly, fast on any machine. This is the only way to have some crisp models, even using 100k quads! This the only way to set-up a pose. This is the only way to have the essential "move" tool. (this exists already in sculpt room). A rigging like method here is always welcome but not crucial.

You all like to work on surface mode. Well, you should do this on a quad mesh, not a tri one.

Many other fixes or new functions are fine but of minor interest IMO.

I'm afraid that Andrew has to write lot of code from scratch. But he is the man for this job. Who else?

In my opinion, these will make 3DC a great app. In a few days new ZB4 is coming, pixosomething bought sculptris too, who knows what else is coming next.

3DC has to be great (it should) and probably more expensive than it is now.

I really like you to comment these, lets stay reasonable, lets not turn this to a wish list, well, as much as we can.

And most of all, please try not to be offensive, I'm not the "black sheep" here.

[Polygoon]

Hi michalis,

In regards to question 1. I think it is always going to be difficult using tri's to quads simply because of the rules of geometry. Not that its impossible I don't know, just that inevitably a model that is made up of tri's is going to have some polygons that can't be quadrangulated (not always). I agree though it would be good if it did.

2. to be honest I'm not sure which is best. I agree that Catmull-clark is good, and the 1 I use most often. But I haven't done much experimenting in this area, so I don;t know.

3. There was a post several months ago that may have gone over similar things. Andrew did mention that bringing over some of the tools from the surface mode to the sculpt room could be done with out too much hassle. He just needed some time to do it, which is perfectly understandable.

Actually there is no big problem. I just need some time to implement what is missed in sculpt room.It can't be done automatically because of very different representations of the mesh in Sculpt and Voxel surface modes.

Not sure whether this is exactly what your hinting at, but here is a link to the post.

My link

Also now that I have been using 3D Coat for a while, I am completely hooked, so I too would gladly pay some extra for it. I think its worth it. But I guess, really that's up to Andrew and Pilgway. They may want to price it so as to ensure maximum chance people will pick up the program give it a whirl, and after trialing it out for a few weeks they go ahead and buy it. The price may effect the number of people that would do this. I don't know. Having said that, speaking for myself, and I would bet that most of the people who already have 3D Coat would gladly pay some more to upgrade to the next version. I for 1 want to help with the development of this great program, if paying some extra to upgrade helps, I gladly will, but for people looking to buy the whole program as new I don't know.

All the best.

[splodge]

i'm just here for the beer.

[michalis]

Thanks for response,

About question 1

What is really happening when re-topo using quads on a voxel model? When subdividing or baking? Voxels produce tri models, working on surface mode can make this clear to you. You sculpt on a tri model. Actually tris is what voxels export. You know this don't you? So its not exactly a question, is where all problems of 3DC exist. Its not about a mac problem, its about a major 3DC issue. PC version has the same problems. I would appreciate it if Andrew could explain this to us here. Why when baking some quads seem like having inverted normals? What's happening when subdividing turns to a mess?

As I said polygoon, its not another wish list here. Its not about the mac version as I feel fine using just 5M voxels with 50 fps. As for the beer splodge, sorry man, I just like this emoticon, no beer today.

[Polygoon]

Yeah, I don't think it is question of where all the problem with 3D Coat lies. I believe everything is functioning how it was intended. Perhaps a triangulation function could be helpful (not a wish, just a thought). By the way, other than quads being needed for Ptex, is there any real advantage to quads verses tris?. I mean if you consider that 2x the polygons can be used for the same number of vertex's if you have tris. Virtually a subdivision for half price, plus the additional angle of edges. Though I am sure there are situations where 1 is more suitable than the other and vise - versa.

[michalis]

A quadrangulated cage, is it possible to capture the voxels to tris model?

It is for blender using Shrink wrap modifier.

I used this voxel sketch here

http://www.3d-coat.com/forum/index.php?showtopic=6091&view=findpost&p=45915

Export a triangulated model from voxels room, a 500K - 1M using decimator is ok. Use export scene! careful here.

Export the quad mesh from retopology room. Don't try to subdivide it, 3DC will mess everything here !

Import these two to blender, In the same position! attention here too.

Subdivide now the quad re-topo mesh to ~ 1M quads using the multi-res modifier. (Catmull-clark method of course)

Shrink wrap it using the name of the tri (from voxels) mesh. Do this for every step of subdivision (its even better and 5 clicks work)

This way we have a subdivided multi-res quad mesh with nice details from the voxel room.

Then:

1. We don't need displacement maps

2. We can make UVs in blender, not in 3DC!

3. We can keep sculpting in surface sculpt room of blender 2.5.3 up to millions of poly

4. We can start rigging using the low poly 2k mesh and set up poses

5. We can bake a perfect AO map or normals or anything for exporting the mesh to another favorite app. Don't be afraid to use tangent nor_maps, blender seams are invisible.

6. We can have multiple textures, nodes etc. We can paint geometrical sensitive patterns on UVs with precision.

7. We can be happy. I used blender 2.49 (a 32bit mac version) every action takes less than two sec! v2.5.3 is 64 bit and even faster. Faster? lol

8. We even can import a low poly version to 3DC and paint some textures (colors of course, not normals)

3DC used here for voxels room and retopo without smoothing and subdivisions.

Conclusion:

Why all these are impossible now for 3DC?

Check out the crucial places behind ears and near openings, snapping in blender is perfect and the method is editable too.

IMO the above is the only clean way to work fast and stable,

A nice way to feel free as an artist.

This is the challenge I'm trying to talk about all these days. This should be the future of 3DC.

Fireworks are great, but...

The magic wang out of the magic castle, in plain sun light, is a wooden stick.

[splodge]

The magic wang out of the magic castle, in plain sun light, is a wooden stick.

[michalis]

Not mine. Wittgenstein wrote this. Sorry, and thanks.

I told you we don't have beers here. Fireworks only.

[artman]

Conclusion:

Why all these are impossible now for 3DC?

Check out the crucial places behind ears and near openings, snapping in blender is perfect and the method is editable too.

IMO the above is the only clean way to work fast and stable,

I got same problem with snapping than you, but I dont need to resnap or relax or subdivide in retopo room.

Retopo room is just the baking station for me.

also I dont experience problems with seams in 3DCoat tangent space normal maps.Its strange that you have.

All my normal maps display without seams in 3DSmax and in Xnormal 3dviewer(I use it as a 3Dviewer for game assets).

But anyway, you seem to have found great workflow combining 3DCoat,Blender and Zbrush.

There is no reason to change it.

On my side it looks like this.

When i want a subdivided quad mesh,I use Merge to Ptex using an assymetric quadrangulated mesh

with quadrangulation settings 1.0 millionpolys/1.0 normalsmoothing.

To get subdivision I use ExportHires...there i get subdivision 4x.16x.64x.

No snapping problem there.The mesh is not conformed,it is displaced.

I get zero details loss up to 8-10 million.Ptex merging is really crisp.

IDEA

Maybe you have wrong export settings for Blender.

There is a lot of possiblities of settings when exporting Tangent space normal maps from 3DCoat

...something like 16 possiblities.

You got TBN calculation method,you have Max/LW/maya setting,and you also have the green channel.

That makes a lot of combinations.

Send a pm to Juan Carlos,he is always using Blender and I think he got good normal map settings.

[Taurus]

3DC has to be great (it should) and probably more expensive than it is now.

You are welcome to pay more I won't my budget is already stretched.

[michalis]

@artman, thanks man, you always helping me, you're right the ptex method can export dense quads without snapping problems. I tested it using my manual quad retopo, I exported something like 200k this way. Some minor problems near openings again but I fixed them in blender (sculpt smooth tool). I can reconstruct subdivisions in zbrush now, or export the low mesh too and construct subdivisions in blender.

But the main reason I started this thread here is to talk about functionalities inside 3DC.

BTW n_maps seams are visible inside 3DC here.