39 replies to this topic

[Cube]

No, voxel count + polygon count affects performance. In Zbrush, you are only worried about polygon count. If you want to compare apples to apples, don't add the voxel overhead, compare teh two when using only polygons. The mesh shown is only a proxy of the underlying levelset surface. The SDK allows you to export this array to your rendering software so you can render the levelset directly. Zero polygons.

- Chad

So using voxels will always be vastly less efficient vs regular polygonal methods in terms of overall geometry achievable on current hardware?

Like my suggestion of an option to remove the polygonal skin from the voxel room and work directly on the levelset (as i understand it there are several alternative methods of smoothing a voxel volume that don't use a polygonal skin), would this be more resource friendly and allow us to work at an additional subdivision level than is currently possible?

This is exactly the stuff i am trying to find out, i realise voxels have many advantages but if we won't be able to get much finer control over surface details at some point i think it's important to know this in advance so people don't get their hopes up too much. If voxels will be limited to mid-high frequency details, with the final touches being done in Paint room or another application then i can live with that, i just want to know one way or the other what's going to be possible in near future with the voxel room.

For me the idea of working on a model from start to finish (including painting when Andrew implements this) entirely in voxels is very exciting, but the overall geometry needed for fine detailing is just not possible at the moment.

[ChadCapeland]

So using voxels will always be vastly less efficient vs regular polygonal methods in terms of overall geometry achievable on current hardware?

No, it depends on the ratio of surface area to volume, and the range of frequencies.

Like my suggestion of an option to remove the polygonal skin from the voxel room and work directly on the levelset (as i understand it there are several alternative methods of smoothing a voxel volume that don't use a polygonal skin), would this be more resource friendly and allow us to work at an additional subdivision level than is currently possible?

You would still need SOME way to rendering the levelset. A raymarcher would work, but it may be slower than polygon skinning. It would be more memory efficient, but slower. Especially if the levelset is not being updated, but the view is changing. So if you are rotating or zooming or whatever, the polygon skin is very fast, since the conversion to polygons from voxels is cached, and the polygon rendering is done in hardware.

This is exactly the stuff i am trying to find out, i realise voxels have many advantages but if we won't be able to get much finer control over surface details at some point i think it's important to know this in advance so people don't get their hopes up too much. If voxels will be limited to mid-high frequency details, with the final touches being done in Paint room or another application then i can live with that, i just want to know one way or the other what's going to be possible in near future with the voxel room.

For me the idea of working on a model from start to finish (including painting when Andrew implements this) entirely in voxels is very exciting, but the overall geometry needed for fine detailing is just not possible at the moment.

There's two options, really. Either break up the voxel array into a hierarchy of some sort or do only low frequency work in voxels and do high frequency work in polygons or displacements. What's the problem with shifting over to polygon surface modeling for the high-frequency detailing?

- Chad

[Cube]

No, it depends on the ratio of surface area to volume, and the range of frequencies.



You would still need SOME way to rendering the levelset. A raymarcher would work, but it may be slower than polygon skinning. It would be more memory efficient, but slower. Especially if the levelset is not being updated, but the view is changing. So if you are rotating or zooming or whatever, the polygon skin is very fast, since the conversion to polygons from voxels is cached, and the polygon rendering is done in hardware.



There's two options, really. Either break up the voxel array into a hierarchy of some sort or do only low frequency work in voxels and do high frequency work in polygons or displacements. What's the problem with shifting over to polygon surface modeling for the high-frequency detailing?

- Chad

I think i understand what you are talking about re: surface to volume, where certain forms will definately benefit voxels, but lets say for example modeling a humanoid figure (probably the most common modeling task in sculpting apps). It seems to me that using a polygonal modeling approach like ZBrush or Mudbox allows for much higher detail levels than we can get with the 'marching cubes' representation of voxels.

Yes i understand that the actual voxel data set still needs to be rendered, i was assuming some kind of cube or spherical primitives for each individual voxel. If you go to the Wikipedia page on Voxels, there is an image on the left hand side of the page which shows what i'm talking about:

http://en.wikipedia.org/wiki/Voxel

Does that type of representation require raycasting? I know that voxels and raycasting apparently work well together (at least much faster than using polygons) but maybe there are other methods that could be used?

There's no problem with using a polygonal modeling app for the fine surface detailing (other than having to invest in more expensive software). That's what i was saying before, if that's the way the workflow has to be for the time being i totally understand, i'm just curious if 3DCoat could really become a one-stop-shop for modeling with some improvements to the voxel room

Perhaps i have come across the wrong way in this thread, but i'd just like to repeat that i'm not attacking 3DCoat in any way - it's awesome software! I'm not accusing Andrews way of working with voxels of being inefficient, he clearly knows a hell of a lot more than i do about any of this! - i'm just wondering if there can be alterations made down the line that will allow for that extra level of final detailing within 3DCoat.

One more thing, there is the possibility of game engines using super high detail voxel volumes for world assets etc.. within the next few years or so (proabably when the next console generation arrives). I know that you could do the lions share of the work in polygons then convert that mesh to a voxel volume (exactly as we can do now in 3DCoat) but there will still be the need to finish of the fine details in voxels before you export to whatever format the engines will be using (SVO seems to be the best bet from what i have read). If that kind of technology takes off then we'll have to get around this fine detail issue one way or another.

I think polygonal modeling tools seem to be at the point of diminishing returns to some extent (ie: when you get to a certain level of detail, the work involved in going much higher does not pay off in terms of what the viewer would actually notice). Maybe it will just be a case of waiting for computer hardware to become powerful enough to allow us to work with larger voxel data sets, so we can match what is being done in ZBrush/Mudbox at the moment, i'd just like to know if there is any room for improvement before that time.

[Colwax]

I'm still very new to this whole process of doing virtual sculpting, so forgive me if my question is not quite right.

If 3D-Coat uses voxels & it appears that there's a problem getting sharp details...
...then what's the difference with the voxels being used by Sensable's FreeForm & ClayTools? (aside from their cost difference to 3DC!!)

Looking at the examples of the models made with FreeForm, it would appear they have no problem getting sharp details...
http://www.sensable.com/industries-design-model.htm

When I bought 3D-Coat, I was really hoping that it's use of voxels would give me a "poor man's" version of ClayTools or FreeForm without the Haptic device.
So now I'm kind of left slightly confused as to why one program with voxels can do this & the other with voxels can't??
Curious to know what the answer might be?

regards Colin

[Cube]

I'm still very new to this whole process of doing virtual sculpting, so forgive me if my question is not quite right.

If 3D-Coat uses voxels & it appears that there's a problem getting sharp details...
...then what's the difference with the voxels being used by Sensable's FreeForm & ClayTools? (aside from their cost difference to 3DC!!)

Looking at the examples of the models made with FreeForm, it would appear they have no problem getting sharp details...
http://www.sensable.com/industries-design-model.htm

When I bought 3D-Coat, I was really hoping that it's use of voxels would give me a "poor man's" version of ClayTools or FreeForm without the Haptic device.
So now I'm kind of left slightly confused as to why one program with voxels can do this & the other with voxels can't??
Curious to know what the answer might be?

regards Colin

Hey don't worry at all, you have made a very good purchase decision with 3DCoat, i've taken a look through the gallery images on the sensable site and 3DCoat can EASILY match that level of detail. I think we just have a different definition of high details.

Just to be clear what i'm talking about in this thread is ULTRA-FINE detailing, the likes of which you can find if you go to the ZBrush website and look through their 'featured' gallery. I'm talking skin pore level of detail here. 3DCoat has really good mid-high detail, there is no issue there at all.

[artman]

Looking at the examples of the models made with FreeForm, it would appear they have no problem getting sharp details...
http://www.sensable.com/industries-design-model.htm

There is not a single model displayed on Sensable gallery that have real highfrequency details.
The detailing capacities displayed on the 3DCoat gallery is of the exact same level.
Of course this level of detail depends on sculpting skill and understanding of resolution vs brush size.
Look at some of Tree321 work or Juan Carlos take on the Invincible character ;those were all made with very early builds of 3DC .
Those guys truly mastered voxel sculpting.
Of course if you increase a sphere to 3mil polys and start goofing around with big brushes you will find it clunky and sluggish.
But if you take a full week to understand how brush size/resolution relation works and you have drawing/sculpting talent you can achieve this level of detail easily
and for less than 300$..
I have a pretty ordinary machine and I get 60fps brush speed with appropriate brushsize on 20 million polys layers,and I can handle 4-5 of those layers (which means approx a 100 mil polys compositions) + I can add a countless amount of smaller res layers (ex: for accessories).
I dont feel any performance lag when I work but again its all a matter of understanding that brushsize/res relation,
because if I want i can also make 3DCoat choke for 3 hours in 20 seconds..

[artman]

Brush size is the least of my concerns when I hit ENTER while in Surface mode, and then when you switch back from Surface to Voxels...maybe you can show a demonstration for everyone to see how we are all missing it but you, because everyone else...including Andrew himself, is aware of the lengthy merge times.

Brush size have DIRECT impact on surface-voxel conversion.
Brush size is what determine the time conversion its gonna take.
First,look at the demon Im working on at page 9 of the 3.2 update thread.
I work ONLY in surface mode.
The torso is 17 million polys.
I switch between Surface mode and Voxel mode every 20 minutes to refresh the topology.
This way I can keep on stretching the mesh forever.
(I dont use Enter thing I use the conversion cube icon...I need all Surface mode brushes :Chisel,Absolute brush ect..)
Conversion time from voxel to surface is ALWAYS below 1 minute.
I switch from vox-to-surf all the time.
It is because I use brush size that I've put as an example in the pic Im taking about
that my conversion time take below 1 minute instead of 3 friggin hours.
(I draw a dot showing the accurate brush size in the pic)
It is accurate brushsize to do detail on a 17 million layer.
If I use a bigger brush size and do even only 5 strokes the conversion time increase exponentially.
But I dont need to because I've efficently used the resolution I had before increasing up to 17 mil...

Look at Tree321 scupts and Juan Carlos work,did they ever whine on this forum about performance?
I guess Im the only idiot around.I think Im gonna spent less time on this forum and more time sculpting.It is surely gonna be a relief to you..

[haikalle]

If you have over 40mil triangles in one layer that is not good. You should split your body into smaller parts Head-Torso-Legs-Arms.
To me your body looks very nice. Ready for Ptex maybe? I don't know a lot about 3d-coat. I just try to help..

[haikalle]

I'm not quite sure but It seems that merging time works exp. to triangles number. Sorry, my english but I hope this table helps:

1 mil = Mergetime * 1
5 mil = Mergetime * 2
10 mil = Mergetime * 5
15 mil = Mergetime * 12
20 mil = Mergetime * 25

Very bad table but I hope you get the point. It is important to keep it in 1-10 range because it's much faster there.

In my worflow I always try to work with 1-5 mil per layer. And when the scuplt is finished then I add them together and
fix the seams.

Sorry Cube, this is off topic...

[artman]

Well...I was just following what both Javis and Andrew advised, regarding the tip to frequently hit ENTER. Are you talking 20 polys or Tri's? What you're reading at the bottom is tri's, no?...at least that's the case in my UI.

20 Tris...3dc indicate polycount as tris.
Whenever you use Enter or Cube Icon...Conversion time is the same.
What matters is 2 things.
-Brush size.
-Number of strokes.
You need to write yourself a chart.
Use a 300 000 voxel sphere in surface mode(clone the layer a few time,its gonna help for testing).
Draw 10 strokes with a small radius with any Sf brushes ...SF.Clay for example.
See how fast is conversion time.
Now increase res.
Do the same thing while gradually increasing brush radius until you see abnormal conversion time.
You will,this way find bottleneck points.
Once that chart is made,you need to live by it.


Here are the rules Ive discovered:
Rule 1#Move brush and big Pose tool operations need to be done at 100000-500000 range.
It's voxels we are talking about...that 500 000 range is your battlefield.
You can do a TREMENDOUS amount of work with it.
All body,muscles,hair mock up,small organic tremulations even wrinckle hinting can be acheived within that range.

Rule 2#Nearly all voxel brushes sucks! Increase and Build brush are by far the worst.
Increase being the worst 3DCoat brush,its only good for cylindrical shapes like fingers,ears rim and it gets slow...before all other brushes.
Airbrush with sharp pen is still pretty cool.
3#Surface mode brushes rocks...especially Sf.Clay ,its Andrew's best brush.

(those are not really RULES,i dont want to sound like I got the truth or something...btw the name artman is based on novel character it has nothing to do with art or anything...but really,its the truth,if you make yourself that chart you will discover that you can comfortably add small hires details to higres layers with small brush size without any problems.)

About your Hobbit:
The body is hires enough now.You need to split the head now...
Anyway that guy is not gonna be naked,right?
I feel the head is two much hires now....maybe its just an impression.
But if the head is more than 1 mil then I think you should start using other brushes than the one you are using..
Once you are really happy with all the the details in the head you can increase to the 8mil-15 mil range.
Once There you only use very small brushes (Sf.Clay,voxel extrude with sharp pen to do wrinckles,Sf pinch to tone winckles ect...)
If brush is small and you got a computer as puny as me Fps and Vox-surf conversion time should be still very cool.
Skin Pores with extrude brush are still gonnaa be possible at a 10-15 mil range but definetely not micropores....
You cannot keep head and torso all together that is for sure.
You are still gonna be able to move nose tip,ear tip ect...but not big changes of head and body with move brush.
All this must have been to your liking earlier...

Ayway,hope that helps you finishing that character..
(3DCoat is not like Zb/Mb...Its more like a special 3D version of Photoshop+Alchemy(Alchemy the soft not the occult art) .)

[haikalle]

Thanks Artman, good points. Also one good trick is to use "Clone and Degrade" option. Let's take a head sculpt for example.
If you have 8-10mil in your head. Then you can try clone and degrade. And compare those two layers together. If they looks
almost the same then you have inc resolution too soon.

[artman]

Maybe Juan is ok waiting hours and hours for a merge. You seem to be on some kind of a crusade here to try and make others, who run into issues, feel like they are just imagining things and that you're the only one here who knows the deepest secrets of the Coat.

No,he wasnt even working in surface mode...all his work is done with the old voxel brushes.
He is a master with extrude brush and sharp pen as you can see on his old man bust.
Really people are not imagining things...they just dont know how to use 3DC ...they try to mimic Zb/Mb workflow
while it require a completely different approach.Even LJB wrote clear post about that and he can do very powerful stuff with voxel sculpting evn with the said brushsize limtations.

And please stop seeing me as someone pretentious who thinks he got the truth about 3DC,its just that really frustrate me seeing how peoples complain on how it cannot do any detailed work while its been proven over and over that its not true.
And Trust me I would really be happy to keep all those "secrets" to myself and just go on sculpting in my corner...(And really sometimes when I see those little arguing,theorizings and neverending feature requests I feel I should just do so...and maybe I will.)but I have this feeling that I know its false,that voxel sculpting is not that weak,and damn look at the gallery...its not just midres stuff we are talking about.Anyway,Im going now..Good night.

[ifxs]

I'm not quite sure but It seems that merging time works exp. to triangles number. Sorry, my english but I hope this table helps:

1 mil = Mergetime * 1
5 mil = Mergetime * 2
10 mil = Mergetime * 5
15 mil = Mergetime * 12
20 mil = Mergetime * 25

Very bad table but I hope you get the point. It is important to keep it in 1-10 range because it's much faster there.

In my worflow I always try to work with 1-5 mil per layer. And when the scuplt is finished then I add them together and
fix the seams.

Sorry Cube, this is off topic...

off topic indeed, but for me and others attempting to achieve even a fraction of the detail that is achievable in nearly every other sculpting app(INCLUDING BLENDER for free), 1-5 million will simply not do. And splitting up a model is absolutely not the way to go for me or anyone coming from mudbox, zbrush or any other "professional" productive sculpting tool; it's bad for baking, bad for uv work(different meshes for the same body/skin), and so much more... IMO its something that has no place in the modern 3d sculpting workflow....

For example, if we're talking about sculpting a nude human figure, why would you ever choose to break up the body parts if the purpose of doing so wasn't for creative reasons? Even with Blender I can get 35million polys moving and sculpting just fine, yes I know the difference between voxels and quad-based poly apps... anyway, 3DC isnt a true-voxel app given the poly-skin you're actually seeing and sculpting.

here comes the reiteration.... :: multithread 3DC, especially the vox room; then merge bars will progress MUCH faster, as will the entire app... And hopefully we will be able to work on high-res detail one day when 3DC has been multithreaded, and either 1 of 2 options that AbnRanger has covered:
#1: we get a multi-res mode in the vox room that allows us to sculpt in a higher level(subD) of detail on our voxel sculpt. Thus having access to all the surface tools for it.
and/or
#2: we get a sculpt room that has a decent toolset, the current one is "a bit sparse" compared to other sculpting toolsets. as per AbnRanger's thorough descriptions

my personal thoughts on the subject.

[ifxs]

Even LJB wrote clear post about that and he can do very powerful stuff with voxel sculpting evn with the said brushsize limtations.

I've read LJB's thread.. and guess what he HAD to do because of the slow speed of 3DC? BREAK UP HIS MODEL, he had to split the head off into another vox tree layer, then smooth out the neck, only to now be left with 2 vox tree layers that are separate, not a good-rule for a productive workflow IMO. ESPECIALLY SINCE 3DC doesn't have the ability to MOVE MULTIPLE LAYERS at once, that is a very key point, that MANY 3DC users recognize.... nothing like the other sculpting apps have to move multiple layers(sub-tools) at once, say with the pose tool... so this workflow that even LJB had to compromise on is not at all keeping with the modern 3d sculpting workflow in the last say 5 years or so...

How very tedious it would be to repose that sculpt layer by layer, or at least a waste of time for a productive workflow IMO? a very good reason ZB introduced transpose master long ago... not a professional workflow at all IMO

[artman]

ESPECIALLY SINCE 3DC doesn't have the ability to MOVE MULTIPLE LAYERS at once

You can move mutilple layers at once ANYTIME with transform tool.
You cannot use Pose tool on multiples layers yet ...
but if what you want is a superior version of Zbrush for half the price with layer texturing capacities then you'll have to be patient a little...
Btw,you've been here for not even 2 months and you've already used the word "multithread" more than a 100 times...I think you should get a tatoo or something...
3DC doesnt have transpose master.
Zb cannot load specular texture or even paint in layers.
3Dcoat bottleneck with brushes happens much earlier.
Zb have a 50 000 polys max capacities for baking comparative normal maps.
3DC cannot create custom brushes.
Zb symmetry is based on he canvas and when you rotate your subtool you're screwed.
3DCoat cannot import giant 30 mil meshes chunk
Zb retopology tools are extremely weak...and you can only bring quady models with barely a few tris..
And so on....
Its all about what you need.
If you need to merge and pose 20 million pieces of mesh,well work in Zb dude...

[haikalle]

I agree. All the points Artman gave, they should be seen everyone who starts this journey with 3d-coat.

[Cube]

Hey i'm glad that this thread at least has people talking but i just want to clarify that this thread IS NOT actually a feature request thread. It's just somewhere to discuss what might be possible in the next 6-12 months or so to bring surface detail on-par with ZBrush. I was hoping that Andrew might come in to explain the advantages and disadvantages of the current 'marching cubes' approach in the voxel room and maybe explain his point of view on what the options might be for greater efficiency.

I'm not suggesting that 3DCoat is flawed or can only do medium frequency detail, it can actually produce some very good details, it's just that last 10% or so where it lags behind for now. Of course the program could stay exactly as it is, and eventually match what ZBrush can do when computer hardware is more powerful in a few years, but if there is the possibility of getting another subdivision level on current hardware with some changes to the rendering approach (or any other method) then i'd be really interested in that!

[ifxs]

yes I meant pose and not transform, my apologies...

ArtMan brings up many valid points. Thanks for sharing your insight/workflow tips.

[Colwax]

Hi Cube,

~
Just to be clear what i'm talking about in this thread is ULTRA-FINE detailing, the likes of which you can find if you go to the ZBrush website and look through their 'featured' gallery. I'm talking skin pore level of detail here. 3DCoat has really good mid-high detail, there is no issue there at all.

Oh I see, "Sharp details means Ultra-Fine details = skin pore detail" Thanks for the clarification.

I'm 50yrs old & only leaned to switch on a computer 5+ yrs ago, all of my 3D experience has only been with CAD-NURBS based programs.
Trying to broaden my horizons with 3D-Coat which means now I'm trying to learn my way through the maze of terms that all of you guys are using here.
Hence my confusion about the "sharp details".

regards Colin

[Cube]

No problem Colin, since we have no exact scale to work from, fine detail is totally subjective and everyone has their own expectations.

ZBrush does have an advantage when you get to extremely high polycounts, but for everything else Voxels are an aweul lot more fun to work with! You should be able to match anything you can find on the Sensable site, i really think the hardware is their big selling point, the software appears to be about the same level as 3DCoat.