create mesh from strands? not extrude along strands

[anhungxadieu]

There's sort of cloth simulation in this movie too. Probably maximum, what Houdini cloth can do (now I couldn't resist...).
Now seriously, cloth sim gives idea for another trick, to use just a plain grid primitive and stick to strands. Where strands are generated as particle trails, with appropriate factory nodes, using fixed strand count. Particles should be generated from curve, attached to moving object, to get an nicely ordered distribution of first row. Modulo formula should be something like point ID of mesh, divided by count of second row, to feed into ID to location. Later, select in array of strand positions (had something similar in some compound, for attaching strand extrusions back to point cloud with strands).
Grid comes later as an target for Syflex Mimic constrain - which could be animated to zero weight in pre roll phase. That is, thing should have a pre roll of at least the same number of frames, as strand trail have segments.

Hi Mathaeus,
Can you explain more about your idea? It's a bit messy for me

Hi rray,
thank for your time to fix my scene , i saw in your fix scene have a points (IDs) multiply with 1000 so waht it mean?
another question , in my scene i attach below it have two pcloud, one for simulation is guides strand and another is fillter strand to create a loft that mean it will more than one set of loft .. that you can see in my image below i want to remove a bridge between each set of strand lofts, how can i do it ...?
thank for your help!
https://drive.google.com/file/d/0Bz5Tvz ... sp=sharing

[Mathaeus]

Hi Mathaeus,
Can you explain more about your idea? It's a bit messy for me

This one is based on 'am Ghosting Trail' compound, I've posted around 2008-2009, and I forgot it . As RRay said, it works *only* if goal is mesh with fixed topology. For thing like in pic, mesh need to be 'ordered' (ID is uniformly distributed along rows and columns) - usually this means just a grid in all 3d app,*maybe* it works for lofts, maybe not. You'll need to checkout this, using ICE visualization or something else.
The rest is modulo/divide/multiply/add trickery, maybe a little bit challenging to visualize in mind, but nothing impossible for everyone who is able to count the numbers. No excuse, I'd say. By the way, whenever you deal with IDs of meshes or strands, there's always something special to do with first or last component in row, some + 1 or -1 or like (71 subdivisions in grid is 72 points).

[rray]

anhungxadieu about the ID*1000, that's the messy trick I talked about.



Using build array from set you get the strandpositions as a long concatenated array.
for example:
strand position arrays of points 1,2 and 3:
[3,4,6,6], [8,7,6,4] and [1,4,6,7] (just integers for this example)

using "build array from set" you get =>
[8,7,6,4,1,4,6,7, 3,4,6,6]

the problem now is that the order is garbled in that array because delete point was used before.

So you do "build array from set" of the IDs then you would get =>
[1,2,0]

which is garbled the same way as the array above

now you get the indices array of the 3 points strand positions =>
[0,1,2,3], [0,1,2,3] and [0,1,2,3]

and then add 1000*ID to that =>
[0,1,2,3] [1000,1001,1002,1003] [2000,2001,2002,2003] .

When you use build array from set on this you get =>
[1000,1001,1002,1003, 2000,2001,2002,2003, 0,1,2,3]

which you can use as the key array to sort the [8,7,6,4,1,4,6,7, 3,4,6,6] array now =>
[3,4,6,6, 8,7,6,4,1,4,6,7]

you could also use select in array instead of sort if you do som pre processing on the numbers

I'll look at your other scene when I have a chancee

[Bullit]

This is turning in a very nice topic

[anhungxadieu]

Hi thank rray and Mathaeus,
Regarding this video, any idea about how they can create uv for that mesh and how about performance i also try to have a long mesh but performance very lag ... i also check this one from Vladimir Jankijevic (movie below) in this movie he created a long strand but only one strand and in my case i need to create bunch of rows of strands (do you think i need to switch from row to column?? ). Any idea to create a system which is stable an fast can be use in production like we want to create a loft through a list of strands.
Thank for your attention,

[rray]

here's some info on the commercial https://groups.google.com/forum/#!searc ... TET08fffYJ

Thanks very much guys, we're glad you like the job!

Some have been scratching their heads about how we've achieved the extrusion effect and creation of UVs, and Manny, you've cracked it already!

After a lot of R&D and smoking heads we've 'settled' with the most simple technique, which seemed to work best to achieve the effect the Director was after, and it also turned out to be robust and fast.

For the majority of cloth elements we've used the following workflow:

After having accurately tracked characters and the camera we've:

- drawn spline curves on a static copy of the character's mesh to define the areas the cloth should be 'emitted' from.
- we then sampled points on these curves, reinterpreted the locations on the animated characters, and
- for every frame of a shot added strand positions, effectively creating strand trail point clouds for the whole length of a shot, for every piece of garment
- the strands were then converted into splines and
- lofted. So for a lot of shots we already head the UVs for free. For quite a few shots, because of the nature of the dancers' movements we needed to re-do the UVs though.

So we ended up with an extruded mesh (one edge loop per frame, or subdivided if more detail was needed) which represented the dancer's movement in 3D space over time.
Using ICE we've triggered vertices at the right time to start simulating (using a Verlet setup) and hid/deleted all polygons 'in front' of a dancer.

For some shots to help joining characters and garment, additional pieces were simulated using nCloth or Marvelous Designer.

That's it, nothing too fancy.

[anhungxadieu]

thanks rray for that info!
look like they use standard loft in softimage!

[anhungxadieu]

hi there,

a little fix so far from my previous post, i still have a problem ... still have to find a way to fix that

[anhungxadieu]

well i fixed it hehe, little clumpy but interesting so far

[anhungxadieu]

so i use that loft into my anime character 可愛いですか。(´・ω・`)
take a look that hair animation in here
https://www.si-community.com/forum/v ... 617#p53617

[forton]

https://www.si-community.com/forum/v ... ude#p28654
This is an icetree that makes extrudes based on 2 curves, I didn't use loops, only modulo and arrays, maybe it's helpfull.

I'm trying to do the same in fabric to extrude from random polygons but it's more like a bunch of code in a node, and very "loop based".

[anhungxadieu]

hi Forton,

Thank you, right now i try to do a uv for the mesh, any idea?

[forton]

Long time ago...

the simplest way is to create a 3D vector array the size of the pointcount, then for each row put values between 0 (begin) and 1 (end) on the first position of the vector (U), same with each collumn to the second position (V), make sure you do it in the same order you created the points, you don't need (W) so leave it at zero.
You can select that array as a UVmap in the rendertree.

You can do it also per node instead of per point, like in my example but that's a bit more complicated.

XSI_SAMPLES Modeling_Cilynder is a good example how to do it with nodes.

[anhungxadieu]

well thank you, i checked it and here is a result sometime it crash like a boss

[forton]

nice!!

Not sure why it would crash, I didn't had many crashes when I did it.
Doublecheck if the pointPos and UV arrays are identical in size, create them at the same moment with the same integer for size (rows * collumns).
don't write values outside the arrays size range.
3 rows of poly's is 4 rows of points etc...

Cheers, Wim

[FXDude]

Hi anhungxadieu,

Having just seen this thread,
http://forums.fabricengine.com/discussi ... ing#latest
I was wondering what currently made (non scripted) mesh generation more or less of a big(ish) deal in FE.

Also did you happen to come across this clip during your experiments?
https://vimeo.com/76854309