Jump to content
3DCoat Forums

V4.1 BETA (experimental 4.1.17D)


Recommended Posts

  • Contributor

Proxy still doesn't work properly even without conversion, it creates tears in the mesh, as shown in the gif, only the new algorithm for liveclay is recreating the missing tris between the edges. It's done live, it's why it doesn't explode anymore. It's still broken. here's what it looks like in zbrush:

brokenmesh.jpg

 

 

As you can see the surface is reconstructed to avoid disconnected pieces, and that's what's creating the mess after conversion, I suppose the conversion function is not understanding the new set of vertices/normals whatever and fills the void in its own way.

Link to comment
Share on other sites

  • Contributor

Proxy still doesn't work properly even without conversion, it creates tears in the mesh, as shown in the gif, only the new algorithm for liveclay is recreating the missing tris between the edges. It's done live, it's why it doesn't explode anymore. It's still broken. here's what it looks like in zbrush:

brokenmesh.jpg

 

 

As you can see the surface is reconstructed to avoid disconnected pieces, and that's what's creating the mess after conversion, I suppose the conversion function is not understanding the new set of vertices/normals whatever and fills the void in its own way.

Thanx :)

edit:What I dont understand is 1) why am I not experiencing this? and also In proxy mode Liveclay is not involved at all,its pure SF mode...so the only thing that remains is weight calculation between the SF proxy and the original LC mesh.So something probably happens at the projection stage that mess up the topology. Again,its just projection its not supposed to change the topology at all...so I dont understand.

 

Andrew?

Link to comment
Share on other sites

  • Contributor

The problem comes from the voxel mode, I always had similar mess as the last step of the gif after a while. Somehow the mesh gets holes, and the voxel mode fills the void how it can (it's the infamous "artefacts" rendering voxel mode useless if you put the crappy brush aside).

This is the same thing with the hole in the sphere gif: bridge creates disconnected pieces along the bridged faces (you can see the running strips in green just before voxel conversion btw !). The voxel mode then tries to fill everything but there's two surfaces, it proceed to fill it in a messy way.

 

 

 

 

What I dont understand is 1) why am I not experiencing this? and also In proxy mode Liveclay is not involved at all,its pure SF mode...so the only thing that remains is weight calculation between the SF proxy and the original LC mesh.So something probably happens at the projection stage that mess up the topology. Again,its just projection its not supposed to change the topology/polycount at all...so I dont understand.

 

 

 

The fixing happens when you brush (smoothing works too as demonstrated on the gif) over the teared part. It's not fixed out of the box, but when you try to overwrite the cells.

 

liveFixing.gif

Link to comment
Share on other sites

  • Contributor

The problem comes from the voxel mode, I always had similar mess as the last step of the gif after a while. Somehow the mesh gets holes, and the voxel mode fills the void how it can (it's the infamous "artefacts" rendering voxel mode useless if you put the crappy brush aside).

This is the same thing with the hole in the sphere gif: bridge creates disconnected pieces along the bridged faces (you can see the running strips in green just before voxel conversion btw !). The voxel mode then tries to fill everything but there's two surfaces, it proceed to fill it in a messy way.

 

 

you said you used Decimate4x....so there is no voxel mode involved...I dont understand.Its all pure SF mode.

The fixing happens when you brush (smoothing works too as demonstrated on the gif) over the teared part. It's not fixed out of the box, but when you try to overwrite the cells.

I dont even get the teared part...

 

edit:but polycount is slightly differnnt when going back, it is not supposed to I think...since surface was only deformed and projected.

Link to comment
Share on other sites

  • Contributor

you said you used Decimate4x....so there is no voxel mode involved...I dont understand.Its all pure SF mode.

 

When in proxy mode yes, but going back from proxy, the way I see it, is VOXEL>surface resampling. Unless you use voxel (reduce) and then it's direct surface resampling and you've no chance of hurting the mesh.

The final mess you see in the gif with the thighs is the conversion kicking in trying to fill the voids of surface, then resampling back to surface mode. Creating the awfull mess.

 

In short: if you've got tears that liveclay didn't fix by "bridging" you're the happy owner of a voxely-mess.

Link to comment
Share on other sites

  • Contributor

When in proxy mode yes, but going back from proxy, the way I see it, is VOXEL>surface resampling. Unless you use voxel (reduce) and then it's direct surface resampling and you've no chance of hurting the mesh.

there is no resampling at all

 

it is all SFmode

1)Original SF/LC mesh (SF mode)

2)-Decimate(decimation IS surface mode operation)

3) projection calculation between Decimated and Original LC mesh.

There is no voxel mode conversion involved when using Decimate4x.

Link to comment
Share on other sites

  • Contributor

there is no resa,pling at all

 

it is all SFmode

1)Original SF/LC mesh (SF mode)

2)-Decimate(decimation IS surface mode operation)

3) projection calculation between Deciamted and Original LC mesh.

There is no voxel mode conversion involved.

 

I'm not sure about that, it may look like it from a user pov but under the hood I suspect voxel operations (step3 for the general shape before reprojection). In any case I exposed why I don't trust proxy, if I don't use it, no problem. Same with copy clay (which imo is a quick fix, it looks like one case of user input not taken into account, I posted on mantis about it) and bridge which creates disconnected pieces.

 

Overall liveclay/removestretching is pretty stable.

Link to comment
Share on other sites

  • Contributor

I'm not sure about that, it may look like it from a user pov but under the hood I suspect voxel operations. In any case I exposed why I don't trust proxy, if I don't use it, no problem. Same with copy clay (which imo is a quick fix, it looks like one case of user input not taken into account, I posted on mantis about it) and bridge which creates disconnected pieces.

 

Overall liveclay/removestretching is pretty stable.

thanx again for testing,Andrew will probably have fun reading this. :)

I still dont understand why we dont get same results at all tough.,,,

 

(as for Copy Clay it  is super unfinished and Andrew warned us about this in last update)

Link to comment
Share on other sites

  • Contributor

I'm not worried about Copy Clay (besides it's super rough in functionality, for instance there's no way to copy a part more than once like I would do with clone stamp; it's clearly a very WIP tool).

 

I've a few questions though, if Andrew would care to answer, related to some function in the software I'm not sure to use correctly.

 

For instance, in voxel menu, "accurate smoothing" is said in the manual to "provides more accurate smoothing at the expense of performances". Is is both in surface or only in voxel ?

I don't see a change in surface.

 

The skip rendering: if I deactivate it there's nothing to see anymore, is it normal or is it the remains of an old functionality long time gone ? In the manual the item doesn't even have an explanation (even the creator doesn't have a clue ? ;) ). And if it works properly, what the use for it ?

 

The incremental render: if I deactivate it there's no changes, it used to refresh the surface only when moving the camera around. Is is useful now that it refresh ALWAYS ?

Link to comment
Share on other sites

  • Reputable Contributor

I had problems using proxy mode similar to what Beat experiences. I do not use proxy mode at the present time.

Linux 4.0.09 non cuda 64 bit Beta.

Edit... I just saw that Linux is updated to 4.0.11. Opps not 11a yet, I had forgotten that I was waiting for 11a so I could test the new brushing changes.

Edited by digman
Link to comment
Share on other sites

  • Advanced Member

Hey Andrew, I hope everything is alright with you guys, I saw that there is some unrest in Kiev lately.

 

Hopefully you guys are able to stay safe.  My prayers are with you guys and your families.

 

-G

  • Like 1
Link to comment
Share on other sites

  • Contributor

I found that I needed to do some boolean operations and so returned a volume to Voxel mode. Having finished and returned to Surface mode, I was disapointed to see that all my painstaking details had become blurred. Once burned, twice warned!

Link to comment
Share on other sites

  • Reputable Contributor

I found that I needed to do some boolean operations and so returned a volume to Voxel mode. Having finished and returned to Surface mode, I was disapointed to see that all my painstaking details had become blurred. Once burned, twice warned!

You can actually do some Boolean ops fairly well in Surface mode, but it's usually somewhat slower. If it's not super dense, try it. Raul did a lot of work with some of them to make the results pretty nice and it wasn't as slow as I thought it would be, the last time I tried

Link to comment
Share on other sites

  • Advanced Member

I found that I needed to do some boolean operations and so returned a volume to Voxel mode. Having finished and returned to Surface mode, I was disapointed to see that all my painstaking details had become blurred. Once burned, twice warned!

Another great reason for a projection algorithm..

Link to comment
Share on other sites

  • Reputable Contributor

I found that I needed to do some boolean operations and so returned a volume to Voxel mode. Having finished and returned to Surface mode, I was disapointed to see that all my painstaking details had become blurred. Once burned, twice warned!

This has been known for a long time... Voxels are even square 3D pixels, no way to have dynamic tessellation... You must increase the overall voxel resolution to retain the details and I mean a tremendous amount of voxel resolution when returning to voxel mode from surface mode depending upon your level of details. It is not burnt once trice shy but understanding the difference between the two, Voxels and polygons...

Edited by digman
Link to comment
Share on other sites

Updated to 4.0.12

- LC-based tools polished. With new fast approach of LC color/specular will be correctly kept during all paint operations.

- several Mantis problems fixed. Especially important fixes:

0001327 "Pick from retopo" loads every single retopo group from the scene

0001336 Vertex Paint Preview Window Bug

0001339 Scene Scale Corrupts Clone Tool

0001077 LiveClay creates holes

Link to comment
Share on other sites

For instance, in voxel menu, "accurate smoothing" is said in the manual to "provides more accurate smoothing at the expense of performances". Is is both in surface or only in voxel ?

I don't see a change in surface.

 

The skip rendering: if I deactivate it there's nothing to see anymore, is it normal or is it the remains of an old functionality long time gone ? In the manual the item doesn't even have an explanation (even the creator doesn't have a clue ? ;) ). And if it works properly, what the use for it ?

 

The incremental render: if I deactivate it there's no changes, it used to refresh the surface only when moving the camera around. Is is useful now that it refresh ALWAYS ?

1) Accurate smoothing - only for voxels

2) Deprecated item, possibly will be removed

3) Incremental render leads to re-rendering only changed part of model if scene is not moved. It speeds us sculpting over highpoly models if you are working over local piece.

Link to comment
Share on other sites

  • Contributor

I still get holes in 4.0.12 using my generalbrush-inflate preset...I attached preset.GeneralInflate.3dcpack

Detail level is at 5 but it is required to get smooth inflating,subdivision is not too crazy as you can see on wireframe,holes should not happens.

post-1195-0-50540700-1387269199_thumb.jp

 

tested on default male bust in DX64cuda

 

I also get weird gathering of heavy tessellated "pimples" when smoothing over tessellated areas (again Im using the attached preset at it seems to be the most problematic one)

post-1195-0-25400900-1387270012_thumb.jp

 

 

Also Relation between detail level and Tesselation seems much smaller is this build and render a lot of my presets useless.

Sharp(creaseclay with sharpalpha and lowspacing)  for example provide very nice subdivision in 4.0.11A at 3.5 detail level but  in 4.0.12 the strokes is ugly and look like a plain surface brush stroke,I need to set detail to 5-6 in order to reach same sharpness as 4.0.11A also the brush feels very different and does not follow surface correctly,it skips some surface when doing not so long strokes....

 

I really cannot work with 4.0.12 so Im reverting back to 4.0.11A until this is fixed. :(

4.0.12 feels faster but 4.0.11A feels much better in term of sculpting feeling.

Link to comment
Share on other sites

Ok Artman, will check what goes wrong, dont worry... and thanks for your feedback!

regarding holes try smoothing or something...are they real? or just flipped faces? or non updated faces? because now there's only one function in code that could introduce holes in theory if goes wrong..but the splitting itself by design is free from holes.

Link to comment
Share on other sites

  • Contributor

Ok Artman, will check what goes wrong, dont worry... and thanks for your feedback!

regarding holes try smoothing or something...are they real? or just flipped faces? or non updated faces? because now there's only one function in code that could introduce holes in theory if goes wrong..but the splitting itself by design is free from holes.

Try it on male bust ,I attached preset in previous page.

Its fairly easy to reproduce within 2-5 minutes max,then you will be able to see exactly what they are.

They dont seem real as you can keep sculpting over them and smooth and they seem to disappear.

 

 

But "pimples" are very real...and nearly impossible to smooth.(you dont see them at first when doing stokes but when you start smoothing you can see them popping...also the wireframe is very clear about this issue)

post-1195-0-94571100-1387278236_thumb.jp

also  easy to reproduce using preset I attached in my last post

(btw I hope you guys are not still testing your changes on friggin spheres :D  )

Male bust is a good test subject because it seems to have problematic topology around nose/eyes area.

Anyway brushes  should work on any topology (where there is no crazy ngons or holes or superweird stuff...of course)

 

 

About other brushes,generally it feels like they are not subdividing enough compared to 11A which imo was just perfect.

Even preset I attached in my previous post is super smooth at 3.0 detail in 11A and is 12 I need to crank it to 5 to get same smoothness.

 

 

I will deep test other brushes and report (I planned on doing work on other stuff today but I will deep test for 1 hour at least)

Link to comment
Share on other sites

  • Contributor

Okay,i attached the presets Im using so you can compare issues Im reporting

artman presets final december.3dcpack

(those are the final version of the older one,they are better imo ,more efficient and straightfoward)

 

 

But really,you dont need my presets... new LC changes seem to not subdivide enough...

 

The whole point is to provide user with brushes that does not require smoothing or global resolution increase to get nice,clean strokes. Im talking strokes that get details ready for baking out of the box on medium res meshes

(100k to 500K).

Reason for this is:

1)smoothing over LC is not always nice and produce all sorts of noisy bumps(same in Sculptris),

2)Increasing the whole mesh resolution just to make previous strokes nicer is absurd and makes meshes too heavy (4X)

 

About specific brushes in 4.0.12:

-Liveclay brush really does not subdivide enough cmpared to 11A.

Try Sci-fi armor 1 and 2 (they are both made of the Liveclay brush) at 4.5 detail level more subd should happens...
Now I need to set it at 6 and its not even that nice...holes happens and not so nice topology also as seen on picture.

again strokes should be much nicer and more subdivided as those brushes are made so that user does not need

to further subdivide the mesh to get clean effect.Straight out of the box nice greeble on 200k meshes (default spheres,default male bust ect...) I also get holes too but at a much lesser incidence than inflate preset.

post-1195-0-18509800-1387279160_thumb.jp

 

Try Broad preset (tube clay)

-same things here...not enough subd...I get holes too but its much harder to reproduce.

 

-I tried Buildclay too...same thing.

 

RemoveStretching seems to be the most reliable,clean and stable aspect of current Surface mode sculpting.

-But I still get crash using Move brush with Remove stretching as reported here:

http://3d-coat.com:8081/mantis/view.php?id=1351

(anyway its probably not related to your changes but to much older issue)

 

I really think with 11A you were really close to perfection, only rare holes happened on some brushes and issue with not being able to do long strokes with some surface skipping as mentioned by Haikale.

 

(also of course there is all the other issues related to proxy mode and BridgeClay found by Beat in previous pages but I'll stick to new LC changes as it is the current issue we are discussing.)

 

Edit:

Also Im a little worried because its not even DEEP testing that Im doing;

it lacks deep interaction with Undo/Redo,Proxy mode,CleanClay,various exports...all voxel menu functions like;Separate disconnect faces,clean surface,clean memory ect...Unfortunately I dont have time to do this.I think you should ask somebody to do that (Daniel maybe?) as it is fairly easy to do,its just about writing down steps to try  and see if it create issues...checking them out as they are verified.

Link to comment
Share on other sites

  • Contributor

Ok...I compared more between build(its hard because Im not supposed to have 2 builds opened at the same time for license reason)

Difference between amount of subdivision is less big in sci-fi 1 and 2 (Liveclay based) then I originally thought it was(although higher values 6-7 are much more noticable in 11A) but difference between strength of subdivision  is clearly noticeable on CreaseClay and InflateGeneralbrush preset.

 

Also the kind of weird subdivisions gathering on the borders of strokes (as picture in previous post) really does not happen a lot in 11A while its a very present issue in 12.Generally 11A feels better and more stable...maybe changes in 12 are too extreme.

Link to comment
Share on other sites

  • Reputable Contributor

Thanks for testing, I found many issues related to what you pointed, fixed almost all now.

Auto Retopo is giving me fits. Don't know what happened. A few builds back, it seemed to work fairly well given the hints you mentioned. But now, nothing I do gives a workable result. The second one was the closest thing that I could get. Had to uncheck Hard Surface, and let 3D Coat do all the guess work. No guides. But the moment I put guides on, it fried 3D Coat's wiring and got abysmal results. To much hit and miss, trial and error, for this to be a usable toolset for the masses.

Link to comment
Share on other sites

  • Contributor

Not dissing on Andrew's effort and talent, but "expected" :)

 

Those would be much more valuable as a cleanup/sculpt process (with reprojection, high five manticor !) than an output one.

 

As for the brushes in 4.0.12, it feels ok to me, maybe crease clay need a bit more division, but it's not a real problem (to me at least).

Link to comment
Share on other sites

  • Contributor
As for the brushes in 4.0.12, it feels ok to me, maybe crease clay need a bit more division, but it's not a real problem (to me at least).

What?! no holes on your side?

 

edit:ok,I remember :) you don't really use any LC brushes beside creaseclay,most of your brushe are SF+RemoveStr

Link to comment
Share on other sites

  • Contributor

Auto Retopo is giving me fits. Don't know what happened. A few builds back, it seemed to work fairly well given the hints you mentioned. But now, nothing I do gives a workable result. The second one was the closest thing that I could get. Had to uncheck Hard Surface, and let 3D Coat do all the guess work. No guides. But the moment I put guides on, it fried 3D Coat's wiring and got abysmal results. To much hit and miss, trial and error, for this to be a usable toolset for the masses.

my experience with it is;

 

1)Try without guides first

2) add guides only where result from 1) is complete garbage (usually 2-3 guides not more)

 

I agree that it would be great that guides and autopo truly work together in order to create a retopology tool with full control over edge placement + automated topology generation.Like a superstroke tool :) . I know Andrew had  completely new ideas regarding this matter when he went in the last stage of autopo improvement...so surely it will pop out sooner or later.

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
 Share

×
×
  • Create New...