Houdini 17

News concerning 3D DCC business
Bullit
Moderator
Posts: 2513
Joined: 24 May 2012, 09:44

Houdini 17

Post by Bullit » 28 Sep 2018, 16:00



Finally a cloth simulation at Houdini level.

User avatar
Draise
Posts: 857
Joined: 09 Oct 2012, 20:48
Skype: ondraise
Location: Colombia

Re: Houdini 17

Post by Draise » 28 Sep 2018, 17:26

I saw this in the Blender Artists forum, this is some incredible next level type of ability.

Bullit
Moderator
Posts: 2513
Joined: 24 May 2012, 09:44

Re: Houdini 17

Post by Bullit » 28 Sep 2018, 20:00

Yes. I just wished they changed their node UI.

Edit: of course there is also the time, resources to achieve that quality level.

User avatar
Mathaeus
Posts: 1763
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Houdini 17

Post by Mathaeus » 30 Sep 2018, 22:41

Draise wrote:
28 Sep 2018, 17:26
I saw this in the Blender Artists forum, this is some incredible next level type of ability.
Now I should admit that few months ago I was exercising the transfers of cloth and soft body sims from Blender to Houdini :). Blender cloth was just a several times faster and easier to tweak, than Houdini ''ability'' to make everything except good looking cloth sim - all that by trying around three different solutions (cloth, grain, wire) some of them helped by GPU, where each one was great and next level in some previous version.
By the way, I love how they are talking about 3-5 faster interaction with growing. From my small knowledge that 3-5 faster could be around 1/3 of some ''normal'' growing with Maya xgen, Yeti or else, that is, current H system is unbelievable slow.
Perhaps they'll have a more of luck this time with that new Vellum thing.

NNois
Posts: 729
Joined: 09 Jun 2009, 20:33

Re: Houdini 17

Post by NNois » 01 Oct 2018, 10:22

Bullit wrote:
28 Sep 2018, 20:00
Yes. I just wished they changed their node UI.
Could you precise what's the problem ? On our side the node UI is by far the best one I can found ;-)

The bad news in H17 is the TOPs and LOPs operators are still in development, I wish they will bring that in H17.5 !

Bullit
Moderator
Posts: 2513
Joined: 24 May 2012, 09:44

Re: Houdini 17

Post by Bullit » 01 Oct 2018, 19:37

Mathaeus wrote:
30 Sep 2018, 22:41
Draise wrote:
28 Sep 2018, 17:26
I saw this in the Blender Artists forum, this is some incredible next level type of ability.
Now I should admit that few months ago I was exercising the transfers of cloth and soft body sims from Blender to Houdini :). Blender cloth was just a several times faster and easier to tweak, than Houdini ''ability'' to make everything except good looking cloth sim - all that by trying around three different solutions (cloth, grain, wire) some of them helped by GPU, where each one was great and next level in some previous version.
By the way, I love how they are talking about 3-5 faster interaction with growing. From my small knowledge that 3-5 faster could be around 1/3 of some ''normal'' growing with Maya xgen, Yeti or else, that is, current H system is unbelievable slow.
Perhaps they'll have a more of luck this time with that new Vellum thing.
Is Blender cloth native or is part of one FEM addon?

Bullit
Moderator
Posts: 2513
Joined: 24 May 2012, 09:44

Re: Houdini 17

Post by Bullit » 01 Oct 2018, 19:39

NNois wrote:
01 Oct 2018, 10:22
Bullit wrote:
28 Sep 2018, 20:00
Yes. I just wished they changed their node UI.
Could you precise what's the problem ? On our side the node UI is by far the best one I can found ;-)

The bad news in H17 is the TOPs and LOPs operators are still in development, I wish they will bring that in H17.5 !
When i tested it was not clear what inputs outputs had available, i prefer ICE node types, i even prefer Xpresso C4D ones.

User avatar
Mathaeus
Posts: 1763
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Houdini 17

Post by Mathaeus » 01 Oct 2018, 20:24

Bullit wrote:
01 Oct 2018, 19:37
Is Blender cloth native or is part of one FEM addon?
It is built-in cloth. I think it's spring-based, at least it behaves like that. If I'm correct, Syflex is also spring-based. There's nothing special with Blender cloth, can't do tearing and such. On positive side, it's relative easy to get believable look, because anisotropic or shearing is described by topology of input mesh, also it's fast.
FEM based cloth, like one in Houdini for now, where simulation is performed on underlying uniform structure and transferred back to mesh, it should be better with collisions and tearing, however it's hard to avoid the uniform, rubber like look. Also it's not fast.

P.S. from what I've read on Blender forums, authors of Next Gen movie did some improvements in Blender native cloth, around performance and behavior, perhaps this is will be included in some next Blender version.

User avatar
mattmos
Posts: 437
Joined: 02 Dec 2009, 16:59

Re: Houdini 17

Post by mattmos » 01 Oct 2018, 21:45

There's a new patch out yesterday or today for blender cloth, haven't had a chance to try it out yet. with that and the new Houdini cloth in 17 i think we finally have some decent alternatives to syflex/ncloth.

NNois
Posts: 729
Joined: 09 Jun 2009, 20:33

Re: Houdini 17

Post by NNois » 01 Oct 2018, 21:55

Bullit wrote:
01 Oct 2018, 19:39
When i tested it was not clear what inputs outputs had available, i prefer ICE node types, i even prefer Xpresso C4D ones.
Haha, that's because you don't understand it the right way.
Think about the execute port of ICE you just need an IN and an OUT, in houdini that's the same, what's available always flows in until you explicitly delete it.

NNois
Posts: 729
Joined: 09 Jun 2009, 20:33

Re: Houdini 17

Post by NNois » 01 Oct 2018, 22:23

... (suite)
I think the big difference with ICE is the data flow through the entire geometry, id/points/polygons/edges/uvs/groups(clusters)/xforms(instances) so a lot, you can't imagine having out port referencing all thoses. The positive point compared to ICE is you never end with the fucking contexts problems and by the years it's a very BIG positive point because you can play with all of thoses data anytime you want with ease

User avatar
Mathaeus
Posts: 1763
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia

Re: Houdini 17

Post by Mathaeus » 02 Oct 2018, 01:44

NNois wrote:
01 Oct 2018, 22:23
... (suite)
I think the big difference with ICE is the data flow through the entire geometry, id/points/polygons/edges/uvs/groups(clusters)/xforms(instances) so a lot, you can't imagine having out port referencing all thoses. The positive point compared to ICE is you never end with the fucking contexts problems and by the years it's a very BIG positive point because you can play with all of thoses data anytime you want with ease
imho it wouldn't hurt for them to go in more abstraction like ICE or Fabric, to be able to build just an abstract array of anything and apply it to anything, points or object transforms, without going with CHOPs. So one would be able to feed something like skin/envelope/bone deform directly by VEX/VOPs, still possible for artist to paint weights. This worked in Fabric for Maya, Fabric was able to feed the matrix inputs of Maya skinCluster node, or array of point positions as well, by choice (and Maya nodes already have all needed inputs for appropriate arrays) . Also to expand (optimized) array operations, to be able to build a feasible custom skin operator in H using VEX. For now, Bone deform SOP is pretty much strange body in H, one has to use special pack-unpack SOPs to easy access the static attributes in VEX, while ( theoretical ) skin operator based on VEX, should evaluate per points loops alive in final stage, which is unacceptable slow (yeah I've tried this).
On other side, ICE automatic context switching, probably will be easy confused by something like Merge SOP in H.
In short, imho it's about design decision, where SideFX choose to do everything 'inside' one object, but not that much ''outside'', neither in completely abstract way. ICE has significantly better connection with deformations typical for character, so we had ICE based Dual Quaternion skin and such. IMHO Fabric had a widest design, but exactly this one is not alive anymore. Somehow it seems that winning design decision is a bit focused (and limited) one, and Houdini choice are effects, that is,what happening with points, voxels and such.

NNois
Posts: 729
Joined: 09 Jun 2009, 20:33

Re: Houdini 17

Post by NNois » 04 Oct 2018, 09:48

The launch event

wesserbro
Posts: 171
Joined: 27 Oct 2012, 18:05

Re: Houdini 17

Post by wesserbro » 05 Oct 2018, 15:31

i see nothing special. just some attempts to catch up with other big players.
well, maybe fluids..

wesserbro
Posts: 171
Joined: 27 Oct 2012, 18:05

Re: Houdini 17

Post by wesserbro » 05 Oct 2018, 15:47

is there anything in Houdini that is robust and fast? I think its always either robust and slow or buggy and slow.

wesserbro
Posts: 171
Joined: 27 Oct 2012, 18:05

Re: Houdini 17

Post by wesserbro » 05 Oct 2018, 15:52

sidefx just got attention (=>money) it cant handle well. houdini has its niche and i dont think it will manage to crawl out of it.