Jump to content
3DCoat Forums

V4.1 BETA (experimental 4.1.17D)


Recommended Posts

  • Advanced Member

 

 

By contrast the decimation worked far far beyond even what 6A could do and it did it fast, like under 5 minutes.

 

http://i.imgur.com/EmaFM1n.jpg

 

http://i.imgur.com/WdR9XcN.jpg

 

http://i.imgur.com/iaa6PC4.jpg

 

 

 

 

 

 

Thats really wow... 

i think im gonna upgrade to V4 and spend some time again in 3dcoat after my long break.

Dont like Manual Retopo, therefore autopo is made for me : :)

 

good work and thanks for sharing  @ ancient Regime :)

Link to comment
Share on other sites

  • Contributor

Ok here is result of this object using 7A:

 

post-1195-0-58390600-1379686495_thumb.jp

 

Im using no harsurface mode,Im aiming at 1700 polys(I know its low value but Im trying to go more toward gamemodel resolution) ,

I use decimation but using such a value that no decimation occur(I dont want it decimated its lowres enough already),

Im also using 1.5 at auto-density:

 

 

Result is much better and much faster than previous build, although  you can see it fails to capture the horn..

With auto-density at zero the horn is not captured either.

Unless a curve is drawn autopo refuse to capture the horn.

post-1195-0-07272000-1379686076_thumb.jp

 

 Using a few  curves  produce EXCEPTIONALLY good results,I mean its 1000x better than what I got in previous build...as good as zremesher quality IMHO and way better than Mudbox automatic retopology.

post-1195-0-33309400-1379687903_thumb.jp

Of course I need to test with more objects but this one is good because its a little organic,a little hardsurface and it has thin parts too.

So congrats! Your efforts are really paying off. :)

 

But is it normal that Quadrangualtion Quality roll out only appears when Hardsurface mode is selected.?

It was not like that in previous build.

 

Unfortunately there is still difference between results if object is a different scale/density...this really need to be fixed.

Longstanding issue:  please add an "ignore previous strokes" checkbox to stroke tool,Im getting crazy undoing accidental connections to backfacing or nearby strokes.Its really annoying issue.

 

also a "no symmetry" option to select is missing,it feels weird to be forced to select an axis even if symmetry is off.

Link to comment
Share on other sites

  • Advanced Member

Ok here is result of this object using 7A:

 

attachicon.gifobject.jpg

 

Im using no harsurface mode,Im aiming at 1700 polys(I know its low value but Im trying to go more toward gamemodel resolution) ,

I use decimation but using such a value that no decimation occur(I dont want it decimated its lowres enough already),

Im also using 1.5 at auto-density:

 

 

Result is much better and much faster than previous build, although  you can see it fails to capture the horn..

With auto-density at zero the horn is not captured either.

Unless a curve is drawn autopo refuse to capture the horn.

attachicon.gifresult 4X.jpg

 

 Using a few  curves  produce EXCEPTIONALLY good results,I mean its 1000x better than what I got in previous build...as good as zremesher quality IMHO and way better than Mudbox automatic retopology.

attachicon.gifwith guides.jpg

Of course I need to test with more objects but this one is good because its a little organic,a little hardsurface and it has thin parts too.

So congrats! Your efforts are really paying off. :)

 

But is it normal that Quadrangualtion Quality roll out only appears when Hardsurface mode is selected.?

It was not like that in previous build.

 

Unfortunately there is still difference between results if object is a different scale/density...this really need to be fixed.

Longstanding issue:  please add an "ignore previous strokes" checkbox to stroke tool,Im getting crazy undoing accidental connections to backfacing or nearby strokes.Its really annoying issue.

 

also a "no symmetry" option to select is missing,it feels weird to be forced to select an axis even if symmetry is off.

 

 

Those are impressive renders. I stuck with the stock 1.2 autodensity.  Big question; did you get the painted on density spots to work without freezing up your process to "Not Responding"?

 

Also you can uncheck the axis and leave them all blank..

 

And just downloaded 7A and now I'm installing. :D

 

http://i.imgur.com/oVuT57p.jpg%C2'>

Edited by L'Ancien Regime
Link to comment
Share on other sites

  • Contributor

Those are impressive renders. I stuck with the stock 1.2 autodensity.  Big question; did you get the painted on density spots to work without freezing up your process to "Not Responding"?

 

Also you can uncheck the axis and leave them all blank..

 

And just downloaded 7A and now I'm installing. :D

Thanx ,I did not thought those were "uncheckable"...would be better if they were checkbox instead of sockets.

Link to comment
Share on other sites

  • Advanced Member

So, installing 7A... :rofl:

 

 

http://i.imgur.com/ZYR0Oa4.jpg

 

 

First attempt with 7A using edge loops and symmetry

 

http://i.imgur.com/5FMXRyK.jpg

 

(Andrew could you increase the desired polycount box back up to 6 figures instead of just 5? Oh wait the boxes DO take 6 figures and more...the box just expands to take whatever integers you throw into it. Great.

 

So it got hung up and stopped responding for a long time (like more than 20 minutes) but hard edge doesn't necessarily work well with organic shapes in the mix

 

http://i.imgur.com/Bf3SblP.jpg

 

Another attempt this time WITHOUT hard surface

 

http://i.imgur.com/Jkf6Whp.jpg

 

I suspect 65000 was simply not enough polygons for this model. On one hand it didn't complete in certain areas but on the other hand the variance in polygon density, without painting any indications of where that should be was outstanding.

 

http://i.imgur.com/pT4qG6P.jpg

http://i.imgur.com/hWHaGMx.jpg

 

 

Still getting the anomalies at the demarcator line of the symmetry

Edited by L'Ancien Regime
Link to comment
Share on other sites

  • Contributor

So, installing 7A... :rofl:

 

 

http://i.imgur.com/ZYR0Oa4.jpg

 

 

First attempt with 7A using edge loops and symmetry

 

http://i.imgur.com/5FMXRyK.jpg

 

(Andrew could you increase the desired polycount box back up to 6 figures instead of just 5? Oh wait the boxes DO take 6 figures and more...the box just expands to take whatever integers you throw into it. Great.

 

So it got hung up and stopped responding for a long time (like more than 20 minutes) but hard edge doesn't necessarily work well with organic shapes in the mix

 

http://i.imgur.com/Bf3SblP.jpg

 

Another attempt this time WITHOUT hard surface

 

http://i.imgur.com/Jkf6Whp.jpg

 

I suspect 65000 was simply not enough polygons for this model. On one hand it didn't complete in certain areas but on the other hand the variance in polygon density, without painting any indications of where that should be was outstanding.

 

http://i.imgur.com/pT4qG6P.jpg

http://i.imgur.com/hWHaGMx.jpg

 

 

Still getting the anomalies at the demarcator line of the symmetry

Try auto density at 0.

and No Tangent smoothing.....

 

Maybe decimation is too strong,,,65k for something with as much thin parts,Im not sure its enough.

Link to comment
Share on other sites

  • Advanced Member

Hard Surface works pretty good with organic objects. I'm just experimenting with different settings here trying to get a grasp of what settings produces what effects

 

http://i.imgur.com/fSooAby.jpg

 

http://i.imgur.com/lKTXVWd.jpg

 

Finding the ideal setting for the big elaborate central frame remains elusive in 7A however...

 

I'm wondering if it's really that important to use decimate here or if it just adds more unnecessary work for the computer to do. That might not be an inconsiderable thing if your autopo computes are freezing up your system into not responding.

Edited by L'Ancien Regime
Link to comment
Share on other sites

Here's a test I did. On the first one (blue) I forgot to turn off Hard Surfaces. The pink one is defaults with Hard Surfaces off. On the yellow one I made density selections around the eyes, mouth, nostrils, and ears. it seems to have ignored this. I also gave it just a few guides one around the mouth, around the eye, the jaw line, then one each for the neck and the shoulder. It seems to have ignored some of these as well.

post-466-0-24830900-1379719344_thumb.jpg

Link to comment
Share on other sites

  • Reputable Contributor

Here's a test I did. On the first one (blue) I forgot to turn off Hard Surfaces. The pink one is defaults with Hard Surfaces off. On the yellow one I made density selections around the eyes, mouth, nostrils, and ears. it seems to have ignored this. I also gave it just a few guides one around the mouth, around the eye, the jaw line, then one each for the neck and the shoulder. It seems to have ignored some of these as well.

Yeah...I tried all night to get something workable on a similar bust. I'm afraid we are just going to have to cede the Auto-Retopo title to ZBrush, in it's recent Remesher. I don't know how they managed to pull it off....but they leaped miles ahead for organic objects, like a character. Especially since they have an Adaptive spacing function/setting that we don't have in Auto-Retopo, currently. Doesn't mean Auto-Retopo in 3D Coat is useless. It looks really good for Hard-Surface. But for Characters....there appears some catching up to do. It doesn't like taking input from strokes, it seems. Cause, whether I add a few or a lot, 3D Coat tends to largely ignore them, and give some strange results.

 

http://www.youtube.com/watch?v=CVoFi2mnyhA

 

http://www.youtube.com/watch?v=ly7WTPG5bAc

 

http://www.youtube.com/watch?v=F7BYTbUIM5I%C2'>

Link to comment
Share on other sites

  • Advanced Member

I seriously wish  Andrew would add a discussion on each of the parameters available on Retopo. Right now we're going on guess work for a lot of things, like for instance the proper crease angle. How should that be chosen, and what consequences will a 57 degree crease angle over a 20 degree crease angle have in the computation of the ultimate result?

 We're just guessing what we're doing here Andrew.

 

 

Also one of the big lessons I'm learning about retopo on some of the biggest scultps, like 6 million polys and over is to not be complacent about the quality of your voxel sculpt. one of the things that tends to happen particularly in big sculpts with really convoluted surfaces is not only bubbles and caves but also weird hiddent excrescences, even when you'd think the surface was smooth. I just discovered one causing a major mess in the generated autopo mesh, a hole in it actually but on closer in spection I found over a dozen of them hidden away on the surface.

 

http://i.imgur.com/WFjWnI7.jpg

http://i.imgur.com/atCpp30.jpg

 

 

So any preliminary to autopo, particularly a big autopo of 50-100 thousand polys will require;

 

1. Close Hulls

2 Fill Voids.

3. Inspect for hidden surface excrescences caused by inadvertant torsion of the surface mesh during the sculpt

4. Experimentation with reducing the overall surface mesh/voxel count by resampling to a lower polycount level if possible though here you will want to consider how a lower poly count in your base voxel mesh might affect the ultimate details in your autopo mesh.. In this case I think I can go from 6 million to 4 million without any loss of ultimate autopo detail

 

Though I must say Resampling itself is probably the greatest cause for the creation of these rough surface anomalies, as I just discovered. Any resampling should be followed by a close scrutiny of your entire voxel surface before autopo is initiated. In a way it seems to me now that resampling before autopo is just a waste of time since autopo is in a sense a kind of resampling anyway.

 

http://i.imgur.com/5Igeyr9.jpg

 

http://i.imgur.com/ZEPEVaq.jpg

 

 

Experimenting with painting on mesh density differential; it seems that 1.4 is almost too much...it really gets the detailed tendrils and extrusions well but it seems to suck the available polygons out of the rest of the mesh.

 

http://i.imgur.com/biDELpy.jpg

 

http://i.imgur.com/19HQ1FD.jpg

 

You know it's really too bad we're still dealing with polygons here...wouldn't it be cool if autopo was doing this with Pixar Open SubD??

Edited by L'Ancien Regime
Link to comment
Share on other sites

Could paint

 

High Density area

or

Low Density area

 

Density area interactive paint vs fixed global density modulator parameter

 

will help too at first stage

----------------

 

Adaptive spacing looks as a triger for high density area autopaint based in angles

Link to comment
Share on other sites

  • Advanced Member

To a large degree the spiraling problem is either eliminated or under control. Even in dense irrational looking meshes on convoluted surfaces where I've come to expect them,  I'm not getting the crazy spirals I was before.

 

http://i.imgur.com/Zn8ByUJ.jpg

 

http://i.imgur.com/5MSBryK.jpg

Link to comment
Share on other sites

Also one of the big lessons I'm learning about retopo on some of the biggest scultps, like 6 million polys and over is to not be complacent about the quality of your voxel sculpt. one of the things that tends to happen particularly in big sculpts with really convoluted surfaces is not only bubbles and caves but also weird hiddent excrescences, even when you'd think the surface was smooth. I just discovered one causing a major mess in the generated autopo mesh, a hole in it actually but on closer in spection I found over a dozen of them hidden away on the surface.

.....

Generally for big and potentially dirty models I recommend the routine:

1) Voxelize and resample to 300-400 k polygons.

2) Fill voids, somooth a little using voxel smoothing.

3) Then run Autopo with decimation.

Link to comment
Share on other sites

While i'm painting depth through Material it's kind of "blocky". Seems like 3D Coat is reducing color depth of texture...

Maybe I'm doing something wrong. I've attached texture, and screenshot whats happening.

Any ideas ?

Is the image 8 bit? 8 bit images are producing such result. Use 16/32 bit images to get good displacement masking.

Or ger image to PS, increase twice, set to 16 bit mode, use gaussian blur a bit, decrease twice and save as tiff.

Link to comment
Share on other sites

  • Reputable Contributor

Yeah...can't get anything going with a human bust/head. No matter what I try, I feel like 3D Coat is fighting me instead of assisting me. ZRemesher works great on a face without even applying guides and density painting. Conversely, here with Auto-Retopo, I can't get a decent result no matter what I try.

 

video of a few different attempts (different settings chosen)

http://www.screencast.com/t/9NkxA1V3nJ

 

ZBrush Z-Remesher:

http://www.youtube.com/watch?v=CVoFi2mnyhA

Edited by AbnRanger
  • Like 1
Link to comment
Share on other sites

  • Reputable Contributor

Wow that is a ton of splines. Doesn't it say to use as few as possible possibly none at all? My test wasn't great but I think I used like 5 of them.

I tried just a few yesterday....spent all evening trying different things. Haven't gotten a single decent result after two full evenings of trying. Tried different busts

Link to comment
Share on other sites

  • Advanced Member

Yeah...can't get anything going with a human bust/head. No matter what I try, I feel like 3D Coat is fighting me instead of assisting me. ZRemesher works great on a face without even applying guides and density painting. Conversely, here with Auto-Retopo, I can't get a decent result no matter what I try.

 

video of a few different attempts (different settings chosen)

http://www.screencast.com/t/9NkxA1V3nJ

 

ZBrush Z-Remesher:

http://www.youtube.com/watch?v=CVoFi2mnyhA

Your autodensity is set at one. I'd set it at 1.2

My understanding is that if you set it at 1.00 then you're saying you don't want any differential in density at all.

It didn't let  me paint on any mesh density on this one..I'll retry it to correct the deficiencies around the ears.

 

http://i.imgur.com/bI0NRnO.jpg

http://i.imgur.com/Icb6wHk.jpg

http://i.imgur.com/NyG4WYd.jpg

http://i.imgur.com/MVpG36m.jpg

 

 

I noticed AbnRanger had holes in his mesh; did he remember to close all hulls and fill all voids before his autopo?

 

Here's the second try this time with painted mesh density difference

 

http://i.imgur.com/v3TFdwg.jpg

http://i.imgur.com/TOdCwzx.jpg

http://i.imgur.com/4knQ3eH.jpg

http://i.imgur.com/kjwsFRo.jpg

http://i.imgur.com/TzwnxRj.jpg

 

The only real flaw was a polygon overlap between the eyebrows. When I get back from the gym I'm going to try a different approach and just paint the entire eyes nose and mouth as one mass of density to see how that works out.

 

First autopo was 80k, second was 65k at 1.24 density differential, and of coure close hulls and fill voids.

 

Also your mesh is set at only 6000? I'd bump that up to 12k or 15k and  have decimate set at 20k

Edited by L'Ancien Regime
Link to comment
Share on other sites

  • Advanced Member

http://i.imgur.com/KwELhWl.jpg

http://i.imgur.com/JpSfSk3.jpg

http://i.imgur.com/hcUgygy.jpg

http://i.imgur.com/31OYOQv.jpg

http://i.imgur.com/2oYeqKi.jpg

http://i.imgur.com/BtHEkpJ.jpg

 

Also when you're painting the density modulator on the model the parameter for this comes stock as 2. In this case in particular but in many other cases as well I think you'll want to dial down this multiplier from 1.6 to 1.2 otherwise the dense area of polygons sucks that polygons out of the sparse areas to such a degree that their meshes become so distorted in those sparse areas that the meshes appear to be tearing themselves apart with long jagged polygons.

Edited by L'Ancien Regime
Link to comment
Share on other sites

  • Advanced Member

OK I watched that Ryan Kingslein video that Abn Ranger posted and gave it another try in the manner of Ryan with Zbrush...

 

http://i.imgur.com/3dQ1tIN.jpg

 

http://i.imgur.com/zkDyh5F.jpg

 

I'm gong to  continue playing with this...I'll ratchet down the number of polygons and even though it's a symmetrical autopo I'll draw full circles around the mouth so they'll go from green to red...maybe that will correct that minor flaw at the corner of her mouth.

 

http://i.imgur.com/x0wu9uR.jpg

http://i.imgur.com/VfnWpuh.jpg

http://i.imgur.com/O4bo1SN.jpg

 

Even Ryan said in his video that the fewer guide lines you used the better....

 

Was it the extra guide lines or the subtraction of 1000 polygons from the sculpt that screwed this one up?

http://i.imgur.com/GSLea49.jpg

http://i.imgur.com/xFifHxh.jpg

http://i.imgur.com/Qg4qpiL.jpg

 

Of course Ryan's sculpt has far more dynamic facial features than mine, so that accounts for part of it, but if Andrew hasn't reached the pitch of perfection that Zbrush has with its remesher, I don't think he's that far off. It's still in Beta after all...Personally I think 7A is looking damned promising.

 

It seems to work better without radial lines, just the loops..plus I boosted the polycount to back up to 7100

 

http://i.imgur.com/LHNnBn4.jpg

http://i.imgur.com/AYF4pLe.jpg

(that's not a hole in the mesh on the side there, that's jsut the voxels poking through;

http://i.imgur.com/N1SpRgF.jpg

Edited by L'Ancien Regime
Link to comment
Share on other sites

  • Advanced Member

So the purpose of the upgrade to retopo was really hard surface auto topologizing; I cut all the organic stuff off my mirror frame and just tried to  autopo it in first Hard Surface and then with identical parameters but not hard surface.

 

 

http://i.imgur.com/7NyGPUW.jpg

http://i.imgur.com/UX6PtbW.jpg

 

 

Without hard surface setting

 

http://i.imgur.com/TsrhOdS.jpg

Link to comment
Share on other sites

  • Member

Is the image 8 bit? 8 bit images are producing such result. Use 16/32 bit images to get good displacement masking.

Or ger image to PS, increase twice, set to 16 bit mode, use gaussian blur a bit, decrease twice and save as tiff.

Thx it worked!!

I tried 16 bit earlier, but with no luck.

I'm using blender for procedural patterns generation. I've generated pattern with 5x Gaussian Smooth  and maximal Anti Aliasing during rendering.

Image saved as 16bit TIFF.

Thanks again :)

 

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
 Share

×
×
  • Create New...