Kristinka Hair 3.0 released

Plugins linking to this thread: (hide)

Kristinka Hair 3.2Author: Anto Matkovic
The July 2014 update is another big one introducing new nodes like »Fit NURBS«, »Curls advanced«, »Curvature Amplifier«, »Grouping in Form« »Follow CurveList« »Stretch Hair« »Hair Filler Rounded« — for a full description see the si-community thread and the updated documentation.

The Kristinka Hair toolset is a new and unique way to set up, style and simulate hair using ICE nodes. A set of fully customizable ICE nodes Scalable, from only a few basic compounds for building basic hair, to very complex structures. Hair styling that always considers the whole shape of the hair. Styling works well for short and for long hair. Unlimited hair length, unlimited number of hair segments. Automatic, procedural generation of details - always with full control. Locks, clumps, curls, turbulence, are created by ICE compounds Additional modifiers, like cutting hairs by external geometry, constant strand length for key frame animation, resampling and subdividing strands, morphing with another hair, modulating hair's distribution over emitter, so user can increase density on most visible areas Full support for the Sofimage's built-in Strand Dynamics Framework simulation engine. Only factory ICE nodes were used, it should work nicely with any Softimage version from 7.01 on.

Other media available: Version 3.1's online 'first steps' tutorial. si-community user Bronco67 has created a video introduction for an earlier version available here. Also: A rendering tests thread, a TV ad by PsyOp featuring Kristinka Hair.

In case you want to apply kH3 nodes and dynamics on top of strands that were not generated by kH3 itself, here is a setup to enable that.

local backup: kristinka_hair_32_02july2014.rar

User avatar
Mathaeus
Posts: 1761
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Kristinka Hair 3.0 released

Post by Mathaeus » 13 Feb 2015, 14:13

julius wrote:thank you for these precisions Mathaeus,
Yeah I went into long stories :) Mentioned tutorial is actually about Syflex connection. Also, as it is noticeable, I'm frequent poster here :), so, if there are questions, I'm waiting...

julius
Posts: 248
Joined: 12 Jun 2009, 10:56

Re: Kristinka Hair 3.0 released

Post by julius » 13 Feb 2015, 14:39

thank you very much ! I'll let you know what's going on ! :-bd

julius
Posts: 248
Joined: 12 Jun 2009, 10:56

Re: Kristinka Hair 3.0 released

Post by julius » 15 Feb 2015, 11:54

ok, i have my first question :

at the 5th stage : you use an "emit extrusion emitter from polymesh". In my node list, I only have "emit MT extrusion emitter", and the inputs are different : I can't connect my polymesh emitter... :-\

User avatar
Mathaeus
Posts: 1761
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Kristinka Hair 3.0 released

Post by Mathaeus » 15 Feb 2015, 19:15

I've checked out the last pack (kristinka_hair_32_02july2014) - and thing with exactly same name is included. Should be under 'stick to mesh' folder, maybe under 'utilities', maybe somewhere else under k Hair 3 master folder (really can't checkout this folder location, just now). About inputs for geometry references, could be light purple input, or dark green one, for appropriate outputs of geometry references ( "value" or "out name", in same colors).

Hope you'll find a way to load and adapt the samples, If you didn't already. With inevitable "I hate to say that", have to say, that in last more than five years, helping through remote instructions in step-by-step style, never worked for more complex setups, like this.

julius
Posts: 248
Joined: 12 Jun 2009, 10:56

Re: Kristinka Hair 3.0 released

Post by julius » 16 Feb 2015, 18:10

ok sorry, I didn't download the latest release :D

it's ok now for the samples too, thanks !

sarah
Posts: 3
Joined: 25 Jan 2015, 08:58

Re: Kristinka Hair 3.0 released

Post by sarah » 12 Jul 2015, 05:30

Just trying my luck to see if anyone had this error pop up in3Delight. "Nan value in geometry of primitive "hair.pointcloud".
I have tried pluging everything off one by one and the only thing that removed this error is swapping the uv of my shaping Nurbs which I dont want to do bc it causes my hair to wrap the wrong way. Render doesnt seem affected at first but seeing as this is supposedly a fatal error Im expecting ill see how these problems are affecting renders at some point.

User avatar
Mathaeus
Posts: 1761
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Kristinka Hair 3.0 released

Post by Mathaeus » 12 Jul 2015, 22:16

sarah wrote:Just trying my luck to see if anyone had this error pop up in3Delight. "Nan value in geometry of primitive "hair.pointcloud".
I have tried pluging everything off one by one and the only thing that removed this error is swapping the uv of my shaping Nurbs which I dont want to do bc it causes my hair to wrap the wrong way. Render doesnt seem affected at first but seeing as this is supposedly a fatal error Im expecting ill see how these problems are affecting renders at some point.
Hello,
Thank you for detailed description. Never sow this message, but this could be just because earlier versions I've used, didn't displayed this.
NaN value means, that 3delight is unable to process some value, supplied by Softimage and ICE. Swapping the uv as a fix, probably means, that there is too small distance between strand segments, somewhere in point cloud. kH just tries to distribute strand segments as much uniformly, but some complex style could change that. It has no built in NaN checking.
Firstly, I'd checkout is it a permanent error - does it repeat if you reload the scene. If so (which is a good sign), there are other ways for reshaping, like changing the count of strand segments, changing the 'first segment' value in emitter node.
When it comes to 'kh follow nurbs' node, parameters like 'blend end' and 'AB end' have significant effect to distribution.
Finally, I'd try to re-parametrize the nurbs surfaces used with 'kh follow nurbs', using uniform or chord length option.

How these problems are affecting renders at some point, depends on 3delight. If it proceeds with rendering, probably it already has a mechanism to fix, maybe just by discarding the corrupted segments or entire strands. In any case, it's better to avoid the NaN thing, in advance.

Toonafish
Posts: 23
Joined: 04 Nov 2011, 10:05

Re: Kristinka Hair 3.0 released

Post by Toonafish » 25 Jul 2015, 01:51

sarah wrote:Just trying my luck to see if anyone had this error pop up in3Delight. "Nan value in geometry of primitive "hair.pointcloud".
I have tried pluging everything off one by one and the only thing that removed this error is swapping the uv of my shaping Nurbs which I dont want to do bc it causes my hair to wrap the wrong way. Render doesnt seem affected at first but seeing as this is supposedly a fatal error Im expecting ill see how these problems are affecting renders at some point.

I recall running into the same thing with Arnold a while ago. Unfortunately Arnold was not as forgiving, but I recall solving the issue by clamping some values in ICE.

User avatar
Mathaeus
Posts: 1761
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Kristinka Hair 3.0 released

Post by Mathaeus » 30 Oct 2015, 00:01

Hello,

just to let you know, I'm slowly working on something close to 'port' of Kristinka Hair to Houdini. Compact port, which means basic set of tools for hair styling, plus animated deformation and connection to external deformers, like H wire simulation. Later, depending of interest, perhaps there will be a room to go wider.
For now, it seems that interaction is comparable to SI, some parts faster, some slower. Just been lucky to deal with pretty much uniform structure of usual hair system, so wasn't so hard to find the shortcuts.
Basic structure is almost the same - emitter, shapers by external geometry, modifiers, own hair filler. This time described by H subnets, still everything in one network, from emitter to rendering - except possible external deformers, of course. (Things in screen shot, that's just a working, decomposed setup).
ICE style of connecting everything to everything won't work in same range, unfortunately. Modulators like 'Strand fCurve' are present, but user is forced to type the name of modulated attribute, in some subnet earlier. Don't like this part, but what's there after three H versions (first time I re assembled it in H 13), it's there.
As usually, everything is based only on factory nodes, so system should be independent on future versions of H. It will not contain any setup for cooperating with built in H hair solutions.

Hope I'll put it on the road before this Christmas. I'd believe already have answers to all questions, just want to have a time to set up it as much simple.

Image

User avatar
Mathaeus
Posts: 1761
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Kristinka Hair 3.0 released

Post by Mathaeus » 09 Nov 2015, 23:40

Just to show some progress. Occasionally happened that design allowed to 'cooperate' with H fur node, too. While thing in screenshot is completely custom, anyway.


Image

User avatar
Mathaeus
Posts: 1761
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Kristinka Hair 3.0 released

Post by Mathaeus » 24 Nov 2015, 02:43

Here's vimeo of first test, of styling over H wire solver - just wanted to see how it's working.
kh_hairsim.jpg
You do not have the required permissions to view the files attached to this post.

anhungxadieu
Posts: 175
Joined: 17 Apr 2014, 10:39
Skype: nguyenvuducthuy

Re: Kristinka Hair 3.0 released

Post by anhungxadieu » 25 Nov 2015, 03:53

wow! it look very nice Mathaeus :x

User avatar
Mathaeus
Posts: 1761
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Kristinka Hair 3.0 released

Post by Mathaeus » 29 Nov 2015, 01:33

Thank you.
So there is another post - process trick. Strands are joined together in post simulation, according to distance, trying to mask the simulated 'stripes'. Smaller distance between stripes is stronger join. Obviously this works better with astronaut suit or some another, more voluminous colliding object. I think there's visible unpleasant interpenetration with shoulder, around third second, or else.
Anyway I think I'll include this option. Still looks better than I've expected, according to movements opposite to original simulation, created by non-simulated joining.

User avatar
Mathaeus
Posts: 1761
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Kristinka Hair 3.0 released

Post by Mathaeus » 20 Dec 2015, 03:10

Hello,
Initial version, that is, created in Houdini Indie and free, is available for download here. For now it is, functionally, almost 1:1 snapshot of latest SI version, minus fur related part, plus some small enhancement here and there. Will see what to do next.

Image

nodeway

Re: Kristinka Hair 3.0 released

Post by nodeway » 20 Dec 2015, 19:02

How fast it is? Do you think it would make sense to port it to C++?

I started converting it completely to VEX. Got 2 nodes mostly remade. It looks that they are at least twice as fast after my tweaks. Didn't tried to use it yet. I'm not sure I follow your logic in couple places there. It would be good to talk about it.

Have you thought about making GIT repository so you could work on it with other people?

BTW. Thanks for sharing.

User avatar
Mathaeus
Posts: 1761
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Kristinka Hair 3.0 released

Post by Mathaeus » 20 Dec 2015, 20:08

mantragora wrote:How fast it is? Do you think it would make sense to port it to C++?

I started converting it completely to VEX. Got 2 nodes mostly remade.
Don't.....
Imho there is no sense to go with C++, as I'm not sure is this a long time configuration. I want to be able to change anything, during any phase of work. Afaik that's what anyone who is able, already doing.
Regarding interaction speed, can't say that much about view port interaction - as H 15 tweak tool has some unique habit to automatically switch back to 'regular' mode if hair curves are behind, and of course the rest of my modeling skills in H are close to zero, so, almost all geometry (nurbs surfaces or curves) is coming from SI - simply it was faster for me to export something and just reload in H. However I think it's possible - this is a hair system, where you able to just set a lower count of hairs and proceed, or to work only on guides. Never heard complains about *that* speed.
About another, imho much more important speed : hopefully, Mantra TFP (time to first pixel) is really nice in H 15 when it comes to polygonal hair, subdividing of curves as well.
About other types of speed:
H XYZ distance VOP together with Nurbs surfaces, behaved much slower than SI Get Closest Location (several times) that's reason to use Raycast as default, which is so so for end user (easy to omit the target).
Loops are faster in H, so, put some additional for Frenet frame calculations.
Hair Filler is sitting on predefined attributes, it's plain interpolation of clones inside triangle, so it's fast by self, both in H and SI.

Note that, all that nurbs stuff isn't something beloved for modelers. There were many requests for polygonal geo, used for all shaping. So, if I were You, I'll try to re assemble something like Maya GMH2, or Max Hair Farm if you're ambitious (while some parts of Hair Farm are copyrighted, afaik).

P.S. regarding GIT and such, don't know how many artists are around it :) Or to say explicitly: I think the current form of download is best to be utilized by someone who want to create and render some hair, *not* to mess with code. If (if.....) I'll see at least five to ten people of this kind, there's sense to proceed with docs, some streamlining and so on. If no (most likely it will be 'no') current form is more or less what I personally need, so doesn't matter, it will stay where it is, already.