V4.5 BETA (experimental)

[Taros]

Looks good but try it with more complex models Carlos or less polys.

[AbnRanger]

It ain't perfect (will require some cleanup work), and it took quite a bit of trial and error, but I finally was able to get a decent result on a head mesh. I know 3D Coat doesn't like many guides, but fewer wasn't helping much. I also cranked up the "Capture Detail" amount to .7

 

 

[AbnRanger]

It also seems to deliver MUCH better results by NOT using the DECIMATE option. But man, is it slow when you don't. We're talking about going to the grocery store and run some errands while it calculates, slow. Wish CUDA (6.5) could be utilized for the calculations and used on the Pose tool, if possible

[Taros]

Looks really nice. We have compared the routines with zRemesher and it seems to be normal today to wait for good results. I am not waiting longer than 2 minutes with my settings and am pleased with the results for now. Even if it would need 10 minutes, it's faster than retopo a complex model by hand.

What are your times Don? Can you post your settings please?

[AbnRanger]

We have compared the routines with zRemesher and it seems to be normal today to wait for good results. I am not waiting longer than 2 minutes with my settings and am pleased with the results for now. Even if it would need 10 minutes, it's faster than retopo a complex model by hand.

What are your times Don? Can you post your settings please?

You can see the settings, here. If I check DECIMATE it is fast (1-2minutes), but the result is very unsatisfactory. But after unchecking that, it took over an hour to calculate, but the result was well within the acceptable/decent range, as you can see. Still plenty of cleanup, but I was surprised. In fact, I was pretty much ready to give up on head meshes and never even bother to attempt them again. But, at least I know it's workable now...just takes some Jedi Mind tricks and a bit of work to get there.

 

Overall, on complex models, like character/creature heads/faces....I'd much rather use the Strokes tool to sketch all the topology > ENTER and be done with it. Much faster than having to make guides anyway and have to jump through so many hoops to get a decent, but not great, result. On secondary objects, I think it is certainly worth using. Just not there yet with more complex objects. It still really struggles in the outer eye region and creates extra geometry that isn't even necessary. You can see what I mean from the chin area, on the example above

[Taros]

Thank you for the feedback. Now I know what you mean. I had the same experience with the calculation times. I thought 3Dcoat crashed and never waited all the time. Instead I searched for a better/faster setting. You seem to have more patience.

 

I agree with you. The Auropo routine is currently nice for all static objects or secondary assets that are not in focus.
My experience is that zRemesher is still a bit better when remeshing characters. The polyflow is better there but not perfect too. Serious artists still retopo their models by hand too.

But we are at a very good point in 3dcoat. Let us wait. Maybe Andrew has some ideas to improve it more.

[Carlosan]

Hi Taros

 

Can i ask a little question ?

 

talking about zremesher routines
- Adaptive size defines a vertex ratio based on the curvature of the mesh.
- To maintain the original mesh shape, ZR may need to increase the topology density in cer­tain areas or produce triangles rather than quad polygons.
 
Zremesher really produce triangles in certain areas ?

 

if yes:

Could 3DC autotopo routines do the same -create triangles- to maintain the original mesh shape ?

 

ty in advance

[Taros]

Hi Taros

 

Can i ask a little question ?

 

talking about zremesher routines

- Adaptive size defines a vertex ratio based on the curvature of the mesh.

- To maintain the original mesh shape, ZR may need to increase the topology density in cer­tain areas or produce triangles rather than quad polygons.

 

Zremesher really produce triangles in certain areas ?

 

if yes:

Could 3DC autotopo routines do the same -create triangles- to maintain the original mesh shape ?

 

ty in advance

I guess it is depending on the high poly template. We have not found any tris in our zR meshes until now.

What I observe is a better overall polyflow. There are some situations, especially in hands where 3D Coat does a much better job. But in faces or elements like ears zRemesher is really good from scratch. I mean without any helper guides.

[Carlosan]

hmmm...

 

let me ask one more

 

overall polyflow... how crease angle option affect it ?

 

Its possible to analyze the curvature -vertex ratio- of the model, before to perform the routine to generate the edgeflow ?

 

ty very much for your time, very appreciated

[Taros]

hmmm...

 

let me ask one more

 

overall polyflow... how crease angle option affect it ?

 

Its possible to analyze the curvature -vertex ratio- of the model, before to perform the routine to generate the edgeflow ?

 

ty very much for your time, very appreciated

No there is no way to analyse it before for the artist. You can set density and guides but zBrush decide itself where more or less polys are generated. But it does it good. 3D Coat does a bit better job sometimes in this field. But as I said: The flow in 3D Coat is not always good for animation. Especially in low poly models.

 

In my eyes both tools are good for secondary models like furniture and or staues. Everything you don't need to animate.

 

We haven't made hardcore tests in zBrush because you don't need them in zRemesher in my opinion. I can ask my friend who is doing the tests for me for making some more. But the zRemesher models were always good.

 

I am preparing a big 3D Coat only autopo test. My new results are very good and I don't need zRemesher anymore... Its in progress and I will publish it in my blog soon.

There are some few rules in 3D Coat. If you know them, you will get very nice and satisfactory results in a good time.

[Taros]

Here we are. My first big Autopo test... It was really hard to wait for all the calculations. But the results are really good.

 

http://3d-coat.com/forum/index.php?showtopic=17997

[L'Ancien Regime]

In the meantime I've made more tests and I think I have very good settings for the current version. Try this, it works well for low poly models too:

Capture details: 50% (The higher your Poly count, the higher details can be. But try it careful and start always at 50%)

Auto density influence modifier: 1

Voxelize Before Quadrangulation: on

Voxelized object poly count: Start at 50% of your high resolution model in rounded thousands. My max value was 3.000 when the mesh began to get worse. My mesh has 5.8 million polys and a value of 2.000 was the best one.

 

Decimate if above: on

Decimate model value: 100

 

Smooth resulting mesh: off

 

No density set needed. No guides needed.

I got really impressive results with this settings.

 

 

 

 

 

Exact same settings only with 10,000 polys.

 

This took about an hour to run through.

 

I'm going to have to try this with Voxelize before Quadrangulation 

Edited July 6, 2015 by L'Ancien Regime

[TimmyZDesign]

I believe that the best method for retopology would be a combination of manual and automatic retopology. It could be called "semi-automatic retopo".

For example, we could create some poly rings and/or poly stars to dictate exact desired edgeflow, and then we could use Raul's new Fill Holes algorithm (in the Retopo Room) to fill in the rest of the polygons automatically.

Essentially Raul's Fill Holes algorithm could be improved to act like a very advanced Bridge Tool, and it would figure out how to make all the necessary connections between the poly rings that we have already established manually on the surface.

I think the problem with the current automatic retopology function is that there is a lack of exact control where edges and polygons will end up. If we could first be allowed to create a simple "frame" on the mesh (for the important polygons), then it would be nice to let 3D-Coat fill in the rest of the "inbetween" polygons for us automatically.

The new Semi-Automatic Retopo algorithm would therefore be a combination of the Fill Holes Retopo Tool and the existing Autopo Tool. It would require a manual setup of a polygon enclosing frame and then we could run the algorithm to fill in the rest for us.

Imagine drawing a ring of polygons around the joining area between an arm and a torso, another ring around the waist, a final ring around the neck, then run the Semiautopo routine by clicking on the torso area, and the algorithm would fill in the entire torso area for you by bridging everything between your three established poly rings, keeping their exact placement and number of polygons intact.

After all, the hardest part of retopology is figuring out how to correctly connect patches of polygons on a surface (by reducing/increasing the number of edges from one patch to another, or altering the flow direction between patches with stars). If those complicated connections could be automated for us then that would save a lot of our time and effort.

I believe having manual, semi-automatic control would be better than changing settings in a fully automatic retopo wizard and running it over and over again, hoping you will eventually get a decent result.

Doesn't anyone else want a fast semi-automatic retopology tool/workflow option like that in 3D-Coat? I think the existing Fill Holes Tool and existing Autopo could be altered/tweaked a bit to make it a reality.

[digman]

Andrew worked on auto retopo but it still needs more time put into the new routine... A few days work or a week's worth is not enough.Of course I have no real information on the time spent accept from the twitter postings and know that you need some days to flesh it out.

 

I commend Andrew for advancing the routine as he said he would after 4.5 was released. This part of 3DC is a very important feature and helps new users to get acceptable results. Also it would speed up static object rendering... Spending enough development time on it is very important and is a great selling point of the software. 

 

Painting density seems to have no effect or still is buggy... 

 

I am impressed with the results at times but still it crashes sometimes, density painting is buggy and sometimes the routine still will produce strange results depending upon the model... 

 

I hope Andrew will not put the new auto retopo routine on the backburner as the former auto retopo routine languished for a very long time, way to long in my opinion...

Edited July 6, 2015 by digman

[AbnRanger]

Andrew worked on auto retopo but it still needs more time put into the new routine... A few days work or a week's worth is not enough.Of course I have no real information on the time spent accept from the twitter postings and know that you need some days to flesh it out.

 

I commend Andrew for advancing the routine as he said he would after 4.5 was released. This part of 3DC is a very important feature and helps new users to get acceptable results. Also it would speed up static object rendering... Spending enough development time on it is very important and is a great selling point of the software. 

 

Painting density seems to have no effect or still is buggy... 

 

I am impressed with the results at times but still it crashes sometimes, density painting is buggy and sometimes the routine still will produce strange results depending upon the model... 

 

I hope Andrew will not put the new auto retopo routine on the backburner as the former auto retopo routine languished for a very long time, way to long in my opinion...

I think simply using CUDA (or OpenCL...so anybody can reap the benefits) to do the calculations could help a LOT. There seemed to be a world of difference by turning OFF Decimate and Voxelize. It takes forever, but somehow it goes from problematic to pretty decent...that is, on a complex object like a head of a character/creature