Page 7 of 12

Re: Kristinka Hair 3.0 released

Posted: 29 Apr 2014, 17:06
by mattmos
Cool thanks for the suggestions, I'll try it out asap.

Cheers!

Re: Kristinka Hair 3.0 released

Posted: 09 May 2014, 14:46
by mattmos
Thanks Mathaeus, removing the point velocity chain has stopped the crash. By removing that does it affect motion blur?

We've got another crash in 2015 when trying to make an eyebrow pointcloud visible, and narrowed it down to using the 'kh3 Deform Hair by Rotation' or the 'kh3 Deform Hair' node. I don't suppose you'd be able to suggest anything regarding those nodes? I can't see anything funky in the nodes themselves - wondering if it is something I need to bring to autodesk's attention?

Thanks again for the help!

Re: Kristinka Hair 3.0 released

Posted: 09 May 2014, 21:30
by Mathaeus
Well this seems to be a first hand evidence. Common thing to these problematic nodes is "Interpret Location to new Geometry". Maybe only this one.
Would be great if you can try, in order:

1:plan B This is a set of setups and nodes, using good old switch context, instead of mentioned node. I've added two versions, one transfers the velocity, second doesn't, just in case it's still something with point velocity. Setup is a bit different - emitter mesh has special version of 'initialize polymesh emitter', having an input for deformed mesh. That is, deformed mesh has to be connected in ICE tree on emitter mesh - not on point cloud, anymore. Everything should behave in same way. Also, added a proper transfer of velocity to strands, in case of connecting to mesh extrusions.
If you're using Nurbs emitter - deformer pair, please let me know to add this option, too (in order to avoid 'interpret' thing, pattern is get.emitlocation > get. pointUV > UVToLocation with deformed nurbs as geo input).

2: optimistic solution - "Interpret Location to new Geometry" in kH stuff, this coming from first phillpack, actually it's not factory node, here. Maybe everything can be fixed by replacing this node in your SI version. Wild guess, there is something not wired properly with my node, or something not considered by SI 2015.

About taking the velocity from deformed meshes, this was somehow problematic. Here it causes slower playback, for example. Some external renderers does not need that attribute, at all. Anyway if this works for you, it works.

By the way, before few days I've discovered that trial of SI 2015 does not exist :D. It seems you're where no kH user has gone before. Any feedback is welcome.

Re: Kristinka Hair 3.0 released

Posted: 09 May 2014, 22:49
by mattmos
Heh, thanks for the plan B, I'll give it a shot. Actually, I hope that the optimistic solution will work...

For the trial you could try this link (from the mailing list):

http://trial.autodesk.com/SWDLDNET4/201 ... lm.sfx.exe

Re: Kristinka Hair 3.0 released

Posted: 12 May 2014, 02:38
by mattmos
Okay so that was a bit too optimistic - the reinterpret location node doesn't seem to be the cause (tried it in a few other scenarios), but rather the pointreferenceframe or polygonreframe nodes. Removing the branches containing those stops the crash, but then the compound has no effect :)

Plan B is working beautifully though - even with the velocity versions! You have my eternal gratitude B-)

The rest of our hairstyles seem to be working just fine, so that's the majority of the work with kristinka 3 holding up well so far. Species, Gear and Redshift are all working as well, so 2015 is not throwing any showstoppers so far.

Re: Kristinka Hair 3.0 released

Posted: 28 Jun 2014, 21:35
by Mathaeus
Hello,

This is screenshot of some new things in upcoming version 3.2. Basically it is a collection of already posted things here on forum, some that flowing around but occasionally are not publicity available, some new. Nothing ground breaking. New nodes should be compatible with 3.1 - that is, should be possible to add some of them to existing ICE tree. I hope I'll put the 3.2 version for download, here, in next few days.

Image

Re: Kristinka Hair 3.0 released

Posted: 28 Jun 2014, 22:16
by rray
Very nice, you've got one of the trickiest hairstyles very natural looking

Re: Kristinka Hair 3.0 released

Posted: 01 Jul 2014, 23:11
by Mathaeus
Hello,

Here's version 3.2. As I mentioned, it's collection of already posted stuff, plus new things.

First of all, there are small workflow helpers, like switches for all node types, in root node. For now, it's possible to separately enable form nodes, enable modifiers, and hair filler. 'Point Color' node is able to assign the random color to components, according to form node ID, chunk or guide.
From improvements, there's option in 'stick to mesh extrusion', to override the local twist of mesh stripe, in favor of 'by walk' defined orientation. That's in case if Syflex simulation creates too wild twisting of mesh stripes.'Deform hair' has both options built in, 'faceted' by polygon orientation, and 'interpolated' which is a mix of polygon orientation and point normal.For now, suggested method for all nodes related to static-deformed pair of something, it is what seems to works in SI 2015 (and everything bellow). Deformed copy has to be connected into ICE tree on emitter mesh, or Nurbs surface. Appropriate nodes have suffix '2'. Like 'deform hair 2'.Everything is supposed to work flawlessly with v3.1 stuff - except the functionality of new nodes, of course.

New nodes:

'Fit Nurbs' 'form' node, as an general styling operator. Opposite to 'follow nurbs', it is for hair that grows perpendicularly to emitter. It's filtered by plain distance to red boundary of Nurbs shaping surface, angle to normal of shaper as well. Should be much easier to figure out the behavior, than with 'follow nurbs'. However, by nature of searching from single line, single node can not cover properly the entire surface of human head, except maybe in layout like in picture. Node has internal trickery, to do proper filtering against self, similar to big 'follow nurbs' (should be visible in pic).'Fit Nurbs One' is a lite version, it takes only one, closest chunk, also it allows multiple NURBS inputs. This one is for individual stripes of hair.

Image

Curls Adv(anced) is supposed to do what ordinary Curls can't - exact 'to angle' blending from root, additional twist of hair stripe, also, what I call 'inverse tip', typical unfinished bending, at tip of curled hair. I've created first version for this pic. Node under the hood, is about ten times more complex than ordinary 'Curls' node. It works in different way, turns are defined by segments, in order to be able to use as much less segments per one turn. It's tied to 'chunk' in kH - one chunk, one curled hair stripe.

Image

More new modifiers, like 'Curvature Amplifier', which exaggerates the hair curvature, by testing the distance between strand segment and midpoint of two neighboring segments.'Bridge strands' which interpolates strand position, to randomly chosen another strand. Finally,'Skip strand', used for detailing in pic in previous post. At some strand position, it draws extensions according to tangents, then it blends the extensions. On flat portion of hair, it does nothing, on curved portion it does deviation, still relative to strand shape.
These three modifiers does not rely on predefined kH orientation vectors, should be easy to adapt them to some other ICE hair system. Also they are based on relative cheap computations, no big deal to use them repeatedly.

Image

There are other news, for example, there's new 'grouping in form' node, allowing to define the group IDs into individual 'follow nurbs' or 'fit nurbs' node. Should be a cleaner solution in case of many 'form' nodes in ICE tree. 'Follow CuveList', is version of 'follow curve', which rely on sub curves into curvelist.'Stretch Hair', more an utility for local re-scaling along strand.'Hair Filler Rounded' distributes strands in one chunk, more in circle like shape, instead plain triangle - it's for use after 'curls advanced' or 'clumps and locks'.

As usual, there are samples for almost all mentioned, docs are updated. This time, there is much more of compounds in pack - for 'just in case' reason, I put the older versions, too. Download ICE compounds, samples, docs.. here..

Good luck !

P.S. From my tests, it seems everything worked smoothly. Anyway don't hesitate to report possible problem - as now we have whole eternity :) for fixes.

Re: Kristinka Hair 3.0 released

Posted: 02 Jul 2014, 13:00
by mattmos
Amazing work as always Mathaeus! ^:)^

The new modifiers look very useful, looking forward to playing with them soon.

Re: Kristinka Hair 3.0 released

Posted: 03 Jul 2014, 03:43
by FXDude
rray wrote:Very nice, you've got one of the trickiest hairstyles very natural looking

It just so happens that I never had to create hair (except long ago using geometry & quickstretch lol)

But this really makes me want to give it a go, from what I can see, it indeed looks like everything is there to make it look just natural!

I'm anxious to see how it behaves in animation, (with collisions, dynamics, etc)
and anxious to shade / render it :)

Re: Kristinka Hair 3.0 released

Posted: 03 Jul 2014, 10:31
by Mathaeus
FXDude wrote:
rray wrote:
It just so happens that I never had to create hair (except long ago using geometry & quickstretch lol)

But this really makes me want to give it a go, from what I can see, it indeed looks like everything is there to make it look just natural!

I'm anxious to see how it behaves in animation, (with collisions, dynamics, etc)
and anxious to shade / render it :)
Hello FXDude
well these questions were discussed before, in many threads on old XSI Base and here. Since first release in May 2009, I think there's total of more than 100K views, of related posts. Anyway it's always nice to see, if someone new is looking at this thing.
Regrading simulation, here's my small test from 2012,wmw of about 5mb, already posted few pages before, of what's called 'stick to mesh extrusion', where actually Syflex is used. There are other methods to simulate, but this one seems to be most popular. Latest version has a few improvements in that method, there is sample with this setup in download.
About rendering, perhaps best known is Psyop's MiO 'Nose Job'. Also I'd be free to invite You to RRay's rendering test thread, or even to my site.

BUT ( there's always some... but), in all these years, the system relied solely on ICE nodes and nature of open, visual programming system. That is, there's always a big chance for new user, to step into a deep hole. No rails around dangerous areas, or at least, not that much.

Re: Kristinka Hair 3.0 released

Posted: 03 Jul 2014, 12:43
by FXDude
Hi, I actually remembered those pics (and also thinking 'wow')

But I admit not having made the association, I actually wasn't really much into any forum t'll about 2012 believe it or not (apart occasional lurking)

But anyhoo, thanks for the tips & those links,
I think it's great for them to be resurfaced again including the nose-job spot,
( making I'm sure not just me go 'wow' once more :] )

Re: Kristinka Hair 3.0 released

Posted: 07 Sep 2014, 02:21
by Mathaeus
Hello,

I think it's easy to see - this is a sort of re-construction of ICE kH styling node called "Follow Nurbs", in Houdini. Got the Indie thing week ago, so this is really just an experiment. Exactly, there's vector surface Operator at bottom of screen, defining the math part of ICE equivalent (if i correctly understood the VOPSOP). And this is the only part, similar to ICE. In short, I think it definitively worth to proceed with hair experiments in Houdini, even I it can lead to something conceptually different. Maybe completely different, maybe not, hard to tell, for now.

Image

Re: Kristinka Hair 3.0 released

Posted: 07 Sep 2014, 15:17
by Tekano
heres a little example of some r&d made from few years ago, melena and kristinka styled hair with custom verlet/melena/spring style forces simulation applied to the guide curves.


Re: Kristinka Hair 3.0 released

Posted: 16 Sep 2014, 01:51
by Mathaeus
Tekano wrote:heres a little example of some r&d made from few years ago, melena and kristinka styled hair with custom verlet/melena/spring style forces simulation applied to the guide curves.
Thanks for posting this.

Back to Houdini hair experiment, just to avoid cross posting, a few early observations about Houdini.

First of all, It's really cool how this app takes the imported geometry, in a way similar to dealing with bitmaps in other 3d apps. That is, imported *obj or *iges still stays referenced on HDD - if I overwrite the file on HDD, Houdini loads the new version. This just calls to do poly modeling, using something else.

Second, Houdini "extract" tool, which creates a sort of SI clone. But, it's possible in Houdini to define on "clone", from which level to extract. Let's say, if I place "null" node somewhere in network ("null" node in Houdini is close to ICE "pass through"), I can call the result of network exactly at point of that "null". And network can include practically the entire scene, allowing to extract, for example, the door from model of house. Well a bit dark side of cool approach is, that it is utilized :), this time for rendering. Mantra likes the procedurally generated hair, at render time (which is also cool) so in this case, an ordinary slapping of hair shader in SI, became interesting network of extracted objects and cross connections in Houdini. Need to say that out of box, from the shelf "add fur", creates all these nodes automatically. Simple shader assignment to polygon geometry could be just a drag and drop. But I wanted to pass the all stages, node by node, so... understanding the rendering setup takes me more time than re-construction of ICE hair styling part. Anyway, once I know how the thing works, that's it.

Here in pics, there's a feature introduced in last Houdini: hair generated at render time (took a few seconds to generate around 50K of long hairs). Finally, new dual-scattering hair shader, providing the complete modern hair shading show. Primary and secondary reflection, transmission, glints. Basically, second part of "dual" does a caustics-like leaking of light, from bright points. Perhaps most visible above the eye. An old school approach was, to provide the transparent shadow shader to hair, which, obviously, does not create enough diversity in coloring, typical for real hair shading. This time I've used what's called "faked caustics". There's 'real' one on menu too, together with a good number of another options.


Image

Re: Kristinka Hair 3.0 released

Posted: 23 Sep 2014, 18:55
by Tekano
hello Matic

am pleased to say am working on a hair job for a commercial here in London and am loving your new KH3 compounds so far. it has bought a lot of disparate elements together and improved considerably the control and styling from when I previously looked at it. I shall endeavour to show and tell when commercial is aired and perhaps a mini tutorial beforehand on dynamics.

just a quick question.

I use many individual follow curve nodes on top of an existing follow nurbs style, but would like some further control on the length of the hairs along each curve. when clump is on or off they are all regular lengths and would be nice to plug an individual KH3 random value into a 'length' slot for each follow curve. is this possible? if I set the length after the follow curves it affects considerably the overall hairstyle and does not compute actual length but a percentage along curve.

thanks again for these great set of hair tools they are proving very robust so far and with redshift its a killer combo