PDA

View Full Version : Sculpted mesh 2 lowpoly base mesh (right way?)


JurajMolcak
02-19-2007, 02:47 PM
Hello folks,

I just installed new Blender release since I was interested in its Sculpt Mode, which looks like pretty nice alternative of Zbrush. Now I have dilema, what is the best / common way of transforming Highres data into base mesh. I found few possible ways, but dunno what could be the best in order to work smoothly. I imported my high sculpted *sphere* into XSI as .obj. But what now?

1. Should I use ultimapper and work with ordinary normalmap plugged into displacement?

2. Should I use parallax solution?

3. Should I use vector displacement way?

I want to focus really on one way, since I need to practise this sculpting stuff. Please, could you give me any suggestions or tips, in order to catch the right way ? Thank you.

CiaranM
02-19-2007, 03:09 PM
1. Should I use ultimapper and work with ordinary normalmap plugged into displacement?

2. Should I use parallax solution?

3. Should I use vector displacement way?




1. You can't plug a normal map into displacement and get correct results. AFAIK ultimapper bakes normal maps, but not displacement maps, which you need. This is part of what Zbrush does well and would be the standard workflow for sculpting -> displacement rendering.

If Blender cannot bake a displacement map for you, then perhaps the vector displacement approach would prove most useful.

ThE_JacO
02-19-2007, 11:50 PM
to my knowledge, ultimapper can do displacement and a few other things beside normal maps.

janimatic
02-20-2007, 09:41 PM
to my knowledge, ultimapper can do displacement and a few other things beside normal maps.
hey, could you please give more details, are you refering to depth maps , does it work with concave/convex parts of high/low meshes ovelapping ?

JurajMolcak
02-20-2007, 11:29 PM
Thanx for your suggestions,
This is very quick "xsi --> blenders sculpt mode --> xsi" pipeline test...
BASE OBJECT:
http://www.dvadetride.sk/Shared/base1.jpg

SCULPTED HIGHPOLY:
http://www.dvadetride.sk/Shared/blender1.jpg

DEPTH MAP (displacement slot) GENERATED BY ULTIMAPER with possibility to add extra details with generated NORMALMAP:
http://www.dvadetride.sk/Shared/ultimap1.jpg

VECTOR DISPLACEMENT (bit longer workflow):
http://www.dvadetride.sk/Shared/vectordisp1.jpg

Looks *not that bad* to me, but is there any way how to "smooth" that visibility of topology (edges) which somehow looks displaced as well? Thanks for your help.

cheers,

ThE_JacO
02-21-2007, 01:05 AM
hey, could you please give more details, are you refering to depth maps , does it work with concave/convex parts of high/low meshes ovelapping ?

just fire it up and check it yourself (if you have a version supporting the feature).
anyway, "all" ultimapper does is taking a mesh, and then from its uv space looks up another mesh. What data you derivate from the look-up is is down to a few options, and it can be normal maps just as it can be a heightmap to displace or bump.

JurajMolcak
02-22-2007, 06:19 AM
Just pointing back my question, coz it is hidden under the images... I am looking for suggestions, how to "smooth" that topology (mesh) visibility which is comming from displacement. You can clearly see horizontal and vertical lines in both ultimaper and vectordisplaced versions. I really canīt see problem, so maybe somebody had this problem before... thanx.

I found, that if I just make displacement from ie. "rock procedural" there is no problem. But greyscale displacement map or vectordisplacement map I am using here are generated out from existing mesh and there is also Tangent Property aplied on low resolution object, so maybe its something with that.

janimatic
02-23-2007, 08:26 PM
just fire it up and check it yourself (if you have a version supporting the feature).
anyway, "all" ultimapper does is taking a mesh, and then from its uv space looks up another mesh. What data you derivate from the look-up is is down to a few options, and it can be normal maps just as it can be a heightmap to displace or bump.
hey ,
well about depth maps, it doesn't solve the problem of concave / convex shapes that you will meet with about any character modeling.
I was wondering if you have a secret shader to solve this issue.
The values should be inverted when the highpoly is outside of the lowpoly.
About Oweron pic, i think you should subdivide the low poly before baking to avoid banding in the map.

ThE_JacO
02-23-2007, 11:11 PM
hey ,
well about depth maps, it doesn't solve the problem of concave / convex shapes that you will meet with about any character modeling.
I was wondering if you have a secret shader to solve this issue.
The values should be inverted when the highpoly is outside of the lowpoly.
About Oweron pic, i think you should subdivide the low poly before baking to avoid banding in the map.

not sure I follow you here.
a displacement map DOES solve concave/convxes shapes just fine, it always had and it always will.
the important thing is that the shapes differentials are suitable to be looked up by the mechanism.

the problem is with things like looking up a bent hook from a flat surface, or other similar problems that would require more then a lookup from a landmark. Such things could only be solved effectively by combining a voxeling of the looked up mesh to solve it, and then some fairly heavy computation to generate a displacement, and a map of disaplacement direction vectors, which is currently not possible in any package I know of to start with (not in one step ala ultimapper at least). You can observe the problem in this thread, with the small hooks on the sides of the center, and the problem there is that no matter what, those shapes simply cannot be described by a heighmap in UVspace alone, no matter how you generate it.

could you post an example of what you I mean? I suspect we're dealing more with language barriers then anything here, as your choice of word describes problem that have been solvable for a long time. a picture is worth a thousand words, give or take 42 words.

janimatic
02-24-2007, 09:28 AM
hey Raf,
Ok some illustrations of the problem.
* 1
First on this model
http://animatic.no-ip.com/projets/vrac/rray_MR_DisplacementHorse.jpg
it worked fine, using a rendermap and following Saturn advise (using a scalar state>raylenght + a raytype switch to mesure lengths) because the model doesn't have concave surfaces.
2 *
Now on that 2nd model this method shows the problems at concave/convex intersection and concave values not inverted (obviously this custom solution just doesn't deal with interpenetrations and inversions....) :
http://animatic.no-ip.com/projets/wiki/incoming/displacementExtractConcaveError.jpg
3 *
Generating a depth map in ultimapper minimise the problem. I am using tangent smoothing at 180, distance to surface 0.1, depth range 0.1 (also the value range make it easier to see in the map)
http://animatic.no-ip.com/projets/wiki/incoming/displacementUltimapExtractConcaveError.jpg
But still in that method, there is a problem in concave shapes if you take a closer look :
http://animatic.no-ip.com/projets/wiki/incoming/displacementUltimExtractResultError.jpg
And the depth map extraction shader being embeded in ultimapper, i guess this can't be solved by the end users, unlike the custom solution from Rray and Saturn.

Anyway it looks like the ultimapper solution bhaves better overall (and is easier to deal with) Maybe the problem is also due to the distance to surface set too high on regions like bent arms where 2 highpoly surfaces are within the same distance to low poly surface.
Plus the texture seams...

Anyway i now understand better your statement : bending of surfaces is causing the problem, not exactly concavity

CGTalk Moderation
02-24-2007, 09:28 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.