Page 5 of 6

Re: Kristinka Hair Tests

Posted: 03 Feb 2011, 09:25
by j3st3r
Guys,

Is it possible to replace hair with polygonal strips? I'm considering to create a polygon hair for game, but I need to replace the strands with poly strips. Any idea?

Re: Kristinka Hair Tests

Posted: 03 Feb 2011, 10:48
by Mathaeus
j3st3r wrote:Guys,

Is it possible to replace hair with polygonal strips? I'm considering to create a polygon hair for game, but I need to replace the strands with poly strips. Any idea?
Of course. Actually about half of development was a prototyping for something like that.
Here is the thread. You'll need some of Helge's hair addons, Melena or MT_strands.

For anything further that I can possibly help, please post in mentioned thread - just to keep everything in one place.

Cheers

Re: Kristinka Hair Tests

Posted: 27 Feb 2011, 06:17
by gustavoeb
Image

I didn't needed such a sofisticated tool for that really, but I was really impressed with what Kristinka can do, and how nice it is to set up. Thanks for the great tool Matkovic!!

Re: Kristinka Hair Tests

Posted: 16 Apr 2011, 09:42
by j3st3r
I've got a problem

When I use turbulence, or curl, the hair render is buggy.

Cheers


Jester

Re: Kristinka Hair Tests

Posted: 16 Apr 2011, 10:13
by j3st3r
And one more question. When I try to use the new ICE modeling feature, nothing happens. I'd like to create extrusion of the the guide curves, but nothing happens...

Re: Kristinka Hair Tests

Posted: 16 Apr 2011, 10:25
by Mathaeus
j3st3r wrote:I've got a problem
When I use turbulence, or curl, the hair render is buggy.
Cheers
Jester
Something like this happens when ICE deals with some very small value - ICE seems to be able to deal with NaNs, but MR not. Latest versions of Kristinka Hair have a few switches for some cases of this, but not for everything, of course.

Don't use these modifiers,that's all what I can say.

Btw, there is a chance I'll post a new version, based on what I have for now - sometimes in (this :) ) summer, maybe. Mainly, that's a cleaning (no 'Follow Curve' anymore, for example) and utilizing the advantages of 3delight, such as automatic curve interpolation. There won't be any testing with MR, probably some of features won't work properly with this render engine.

Cheers

for another question, I don't know - maybe some of attributes is missing.

In short, use Melena.

Re: Kristinka Hair Tests

Posted: 17 Apr 2011, 10:05
by j3st3r
Sad news, friend. To be honest, as I mentioned earlier in a discussion, I'd like to use both, both has it's strength. At this moment I need to create hair, that can be baked onto a low poly surface. Or, if I could do a good geometry representation, that'd be the best. My ICE knowledge is far from the enough to resolve this.
First I wanted create MT_Extrusion. With strips, I was close enough, but I need a tapering effect but I couldn't reolve it.
I tried to to use the new ICE modeling feature, but for some reason the polygons created with the new extrusion along strands are placed in the origin, not taking the form of the guide strands...

Unfortunately the task is to make believable hair with Softimage, similar to the Samaritan video hair...Polycount (almost) doesn't matter

Thanks


PS. By the way do not abandon MR support please, not everyone has access to 3delight, Arnold, and so on, please

Re: Kristinka Hair Tests

Posted: 17 Apr 2011, 13:26
by Mathaeus
First of all, 2 core version of 3delight For Softimage is free. How fast it is, when comes to hair rendering, you'll need to try.

Problem exactly begins from kH2 Emit Hair node, when it tries to create barycentric coordinates for guides/hair interpolation, performing a few simple math calculations on
very short distances, between hair/guides roots. In latest Kristinka Hair (now I *realy* hope you're using the latest one), there are a two simple methods to avoid this:
1: sampled root positions were slightly 'pushed' along emitter normal.
2: in case the hair root can't find all three guide roots, needed for 'even' interpolation, compound will switch to another interpolation type (just 'extrusion' along single guide). This can happen with a very low counts of guides, also some modifiers such as kH2 Follow Curve, are able to force that.
From my knowledge, these methods are usualy enough.

AFAIK, Melena doesn't use this way of hair filling. I think it's a random spread around guides, instead of searching the guides in another point cloud and creating the barycentric cords. So, temporal solution in Kristinka Hair would be to just avoid the second, filler point cloud. I think a lot of people are doing exactly that. Disadvantage is a much slower FPS, advantage is a much more predictable styling.


Now, a little surprise. In image bellow, I've intentionally disabled the first method (otherwise, everything is OK here, in Mental Ray too). As you can notice, 3delight still does the job as you expect ! You'll also notice both diffuse colors from hair shader - in Mental Ray image, that's only 'B' color.

Image

AFAIK, it takes about few weeks for 3delight guys to make 'ground' for ICE strand rendering (counting from first announce on forum, and I'm pretty sure this wasn't their only job). Also, I've been able to finnaly figure out some very important facts about motion blur (for me, artist), just from a few short answers by their team lead.

Now, is it a Mental Ray issue or Softimage issue, I really don't know and I don't care of. However, as an artist, I'm 101% sure that the way of developing and distribution of Mental Ray, is not suitable for such complex product as modern renderer is. Probably a gamer graphic cards can go in this way of mass production - if some glow effect doesn't work in your game, you'll switch it off and you're OK. But the renderer simply needs more love, that only small, dedicated team can put into product. It's not only about 3delight. Max's users will know, if Normal Map shader in Max doesn't work well with Vray, they will not wait, they will write a special Normal Map shader for Vray.

It's not that I don't want to support this or that, I just have not a time for doing a nasty workarounds for each version of SI/MR.

Cheers

btw, If you aren't absolutely sure about all mechanisms inside the ICE compounds, please do not mix Melena and Kristinka Hair.

Re: Kristinka Hair Tests

Posted: 19 Apr 2011, 13:53
by j3st3r
Thanks Matheus,

I see your point, and agree with it.

I didn't intended to mix compounds of Kristinka and Melena, but use a mixture of Melena Hair and Krisitinka Hair...both has advantages. Unfortunately I couldn't bake hair in MR, so my project is stuck...

Anyhow, just aquestion, if you have access to SI 2012, do you know why the new extrusion along strands doesn't make the extrusion with Kristinka Guides?


Thanks



Jester



PS. I'll check the 3delight, because I'm also frustrated by MR many times...That's why I gave up rendering a time ago...

Re: Kristinka Hair Tests

Posted: 19 Apr 2011, 14:53
by Mathaeus
as soon I'll have a time to play with trial, I'll checkout the extrusion stuff.
In the meantime, I think it may be one or both of two:
1:
some attributes are missing, as kH2 Emit Guides sets only necessary. You can take look into ICE tree that generates the extrusions correctly, then find in emitting node, and/or strand creation node, what attributes it sets. In 'emitting' node, that might be 'init particle data' node or something (copy this one and plug it after the kH2 Emit Guides). Then you can try to set another attributes from strand creation node (Self.StrandThis, Self.StrandThat...).
Two very important attributes for proper extrusion, strand up vector and strand tangent, are defined as custom in Kristinka Hair (ironically, in order to make the connection with old MT extrusion, to prevent the ICE to kill it's 'own' attributes).
So at the end of ICE tree (exactly, after the last 'deforming' node), you can set:
Get Data> Self.kH_UpVectorAlongStrand > Set Data> Self.StrandUpVector
Get Data> Self.kH_StrandTangent > Set Data> Self.StrandTangent
also...
Get Data> Self.StrandPosition > Set Data > Self.StrandCenter

.... or anything of factory stuff that sounds like this.

2: Point Cloud created in previous version of SI (I don't believe this may be a problem). Anyway, you can try to create a new point cloud, then just copy all the nodes form old one. Easy way for that: in 'old' point cloud, plug everything into execute node, instead directly into main ICE port. RMB on execute node, copy, paste into the new point cloud.

Cheers

Re: Kristinka Hair Tests

Posted: 22 Apr 2011, 16:35
by j3st3r
Matheus,


Do you have any idea if 3delight can bake hair to texture? In game developement it's very useful feature...


Thanks



Jester

Re: Kristinka Hair Tests

Posted: 22 Apr 2011, 17:40
by Mathaeus
j3st3r wrote: Do you have any idea if 3delight can bake hair to texture? In game developement it's very useful feature...
If you mean finding a farthest strand/hair segment, looking from surface of hair emitter (or something close to this shape), along surface normal - and storing the distance in bitmap...
Hardly I could imagine this in out-of-the-box 3delight for SI plugin.

Anyway, if default, Mental Ray render-mapping doesn't work correctly if you're using a long distance from surface (that is, tracing a ray 'back' to surface, along surface normal)... probably you can simplify process (to SI, not to user :) ), by just pushing the copy of mesh to cover the hair, inverting normal, then render-mapping with AO, using a very low angle (or zero) for spread. Then you'll need a color_math_basic node to subtract result from 'push distance', probably some change_range to fit all into 0-1 range, if it's needed.
Honestly I didn't tried all what I said, but I believe this should work.

Btw I could imagine an ICE way for this, but I believe the render-mapping (with MR) should be an easy way.

Re: Kristinka Hair Tests

Posted: 06 May 2011, 22:09
by Mathaeus
Regarding strand extrusions, I just sow the answer on SI mail list, from mr Laforge:

'If the extrusion is not generated it could be for two reasons.
"Create Extrusion Along Strands" doesn't need some orientation attributes on the point cloud. That's because it does its own calculation to find the good orientation. To do this, it is using the point position and the first strand position as a tangent basis. So maybe Melena system create the first strand position on the point position ? If so, the extrusion will not work. What you could try is doing a remove form array on strand position to remove the first element (as it is not needed in fact).
Also, if Melena create strands with some non constant number of positions (like 10 strand position on point 0 and 15 on point 1), it could give some trouble.'

So it might work with Kristinka Hair, theoretically. Mentioned 'first strand position on the point position' is not default option, even it's possible to enable it with 'Start From Root' in kH Emit Guides node. Also, strand positions are 'non-constant' only if you're using 'kH Optimize' node.
Anyway, mentioned method of calculating the rotations (I think it's similar to 'Deform by Curve'), IMHO, is not the best idea for any decent hair styling, as just the first tangent constantly changes. Probably this looks like good idea in first two days.

So, if there is compiled version of Melena, I would go with extrusion operator from Melena. Btw, Helge gently left the up vector as an open option, and adaptation of 'Prepare for extrusion' node that you'll find in KH pack, automatically takes already created up vectors.

Also, if you aren't ICE enthusiast or TD, and you can afford another 3d software, probably you can go with professional solution: Hair Farm plugin for 3d Studio Max. AFAIK, anyone with these two 'qualifications', already did that.

cheers

Re: Kristinka Hair Tests

Posted: 08 May 2011, 03:19
by guillaume
Hi all,

I'm currently looking at the "Create Extrusion Along Strands" problems.

It looks like it works correctly with Krinstinka hair system ...most of the time. I've got some artifact on the roots of the meshed strands sometime but I found the problem. I'm also looking at making it working correctly with Fabricio excellent Strand Tree compounds. Those problems should be fixed very soon, so I will send you a link to a new version ! I couldn't find time to test non-factory compounds before, but I just wanted to say that i will do my best to support compatibility with those public compounds !

Cheers,

Guillaume

Re: Kristinka Hair Tests

Posted: 09 May 2011, 01:04
by Mathaeus
Thank You kindly! Of course, I can adapt Kristinka Hair too. Unfortunately I still didn't had a chance to play with SI 2012, otherwise I would try to do something... Any info is helpful !

cheers

Re: Kristinka Hair Tests

Posted: 09 May 2011, 17:18
by guillaume
You are welcome Anto :) .

Here is "Create Extrusion Along Strands" V2 working with your hair system (and also Fabricio procedural Tree) : viewtopic.php?f=15&t=1456