Wire capture, controlling width per point?

[martinT]

Hi!

I'm wondering how to control the width of the wire captures envelope, per point preferably.

I've read the documentation but I don't really understand it: 

Quote

If the second component (Lookup) contains a channel, then it is evaluated using the current u value of the primitive instead of time. 

Does this mean that we cant use any type of attributes to drive the capture radius?

I would have hoped it would be as easy as with the poly wire sop, where we just create a width attrib and point to that in the sop, but that does not seem to work for me.

Regards,

Martin

 

[Sepu]

What about just using a ramp to shape the points. it will give you way more control. You use a wrangle like this 

float shape = @ptnum / float(@numpt);

@width = chramp("myramp", shape) / 10;

 or you can also put a UV texture, with attrib class to pts and in a wrangle

float shape = @uv.z;

@width = chramp("myramp", shape) / 10;

 

rampPts.hip

Edited November 11, 2017 by Sepu

[MrScienceOfficer]

Wirecapture uses a keyframes on the radius and lookup parm to drive the values by essentially hacking the animation editor instead of using a ramp parm, I think it's an outdated workflow.  I wouldn't recommend using it.

wire_capture.hiplc

[martinT]

Thanks to both of you!

In this case the problem is only with the wire capture node and not setting the width of the poly wire.

What I'm trying to do is rig a speed tree mesh for dynamics in houdini. So I've got curves from speed tree which has resonable width attributes already, and those curves play nicely with the wire solver. So the problem for me right now is to get that motion back to the mesh, the wirecapture/deform seemed like my best bet, but maybe not? Maybe lattice sop? or transfer attribs some how?

Thanks for the example file Tom, it doesnt look like that approach will work for hundreds of branches, or am I missing something?

[Noobini]

just a long shot...but the new edgetransport (16.5) has a Network option for tree/branches structure...have you taken a look and see if it's usable in this case ?

[MrScienceOfficer]

I think you would need to go through all the curves in a for loop and drive the widths by pushing the attributes through CHOPs, it sounds a bit convoluted, and I imagine there must be a simpler way.

Using point deforms is another choice, but you can't, as far as I know, control the capture radius per point.

A third option would be to generate bones from the curves and use the biharmonic capture SOP to bind it.  This would likely result in the highest quality(and most expensive) results since it will use proper transforms to drive the tree, while the other methods work a bit differently and can have less stable results (under more extreme deformation).

I think your best bet though, would be to post this on the SideFX forum as well there's a good chance some one out there has already solved this problem.