Berry like object

[Pancho]

I want to model these sweets. Kind of a raspberry made out of sugar and a wine gum like filling which makes the sugar balls stick to it.

 

The way the sugar balls are packed is incredible. Almost no gaps whereever you are looking.

 

O.k., I made my base object, scattered 210 points (yes, I counted them) and here the trouble begins. In order to have a smooth relaxed point distribution I use relax iterations. Unfortunately the relaxation doesn't take into account the length of the surface between the points, but just the distance inbetween points. This results in a gap between the bottom and side polygons.

 

One idea how to to fix this is to convert the raspberry into a sphere for even distribution of points. But how would I move the points from a spheric raspberry to a regular one? In SI there is "Reinterpret location on new geometry". No idea how to achieve this in Houdini. The idea would be to have two identical models, one deformed, and use the deformation difference to move the points.

 

Also, the scatter node always uses the same radius. In my case the size of the sugar pearls differs. This wouls also be a challange.

 

So, in order to close the gaps and make them fit nicely I thought about a jelly core which attracts the pearls with a force while the pearls expand themselves to their full size. The pearls would behave like small balloons, pushing each other apart while trying to stick to the surface.

 

This technique would be even more interesting for modeling a real raspberry. But I have no idea either how to do this in Houdini and whether cloth or FEM would be the way to go.

 

So, I'd be really glad if some Houdini pros could point me in the right direction.

 

Cheers and thanks in advance

 

Tom

 

 

 

BTW, since I counted all pearls and pealed them off, there is only one layer of pearls.

[JohnMarston]

They're actually arranged via a phyllotaxis pattern, common in nature: https://en.wikipedia.org/wiki/Phyllotaxis

 

A search on vimeo revealed this:

 

So it's a bit simpler than you thought. Thanks nature!

 

Should get you started in the right direction.

[Nico]

Hi Tom,

 

to answer how to  "Reinterpret location on new geometry" in Houdini:

You should activate "Prim Num Attribute" and "Prim UVW Attribute" in Outpute Attributes Tab. Then you use a attribute interpolate sop to move your scatter points with the deformed geometry. The help file of the scatter node has a good explanation about this.

 

And if you activate the "Output Radius Attribute" and paint or use a texture to define "density" on your geometry than the pscale of the points will varies accordingly your density values.

 

Greets

Nico

Edited March 3, 2016 by Nico

[Pancho]

That's quite interesting! Thanks for posting this.

 

Unfortunately the video misses exactly what I need, a way to place the spheres tightly together, with different radii and without interestions.

Edited March 3, 2016 by Pancho

[JohnMarston]

The point relax node allows one to pipe in a pscale so points of differing scale would be repelled with respect to that. I do think that your best bet would be the phyllotaxis though - by nature, at ~137.5 degrees it produces the best packing algorithm. Then you can just vary the size by a bit with some noise or pscale.

[fredhopp]

I hope this setup can give you ideas. It worked for me. I deleted the file caches between the different steps, so it's going to take about 2 min to cook

raspberry.hip

[Pancho]

Thanks, Nico! That worked.

 

But comparing my results with the original sweet don't make me too happy. I'd like to try and go the dynamic way and combine an rbd with softbody dynamics. RBDs for the assembly around the jelly core and softbodies in order to make the sugar pearls fit better together. Looking at the pearls with a microscope reveals that they are hardly round all together but deformed. I would like to know how they manufacture this stuff in reality.

 

Are there any know examples of combining both dynamics? I'll take a look into the H examples, but am glad for any hints.

 

Cheers

[JohnMarston]

@fredhopp - that's beautiful! thanks for sharing, a lot of cool techniques in there.

[Pancho]

Thanks soooo much, Fred!

 

So much stuff to get ideas from! Great for learning!

 

Very generous!

 

Cheers

Tom

[Pancho]

O.k., took me some time, but finally I managed to create pearls which self-inflate and get attracted by the jelly.

 

The only paint at this point are the simulation times. One a i7 six-core it takes 30-40min to simulate about 75-90 frames. That seems totally ridiculous to me. So, I wonder whether there are some settings which are wrong or if I didn't include something which is needed. The parameter in my sopsolver is v which gets altered. One attribVop adds the attraction force, the other the inflate force. Is v the wrong parameter?

 

Would be great if somebody who, contrary to me, knows his way around this kind of stuff would take a look.

 

Cheers

Tom

berry_inflate_01_f_001.hiplc

[Pancho]

Quick Redshift render test. Not 100% there, but quite good...

Edited March 8, 2016 by Pancho

[6ril]

Hi Pancho,

 

I've opened your file, out of curiosity as I'm only learning Houdini, and won't pretend to provide help.

But, as you seem to be aware, the cooking time is way to long for a relatively simple object.

I think your approach is too complicated, look at all the solvers you've got in your dop

Also, the way you create the middle part, with all your smoothing with foreach seems way too complicated! Why not modeling this part in a more simple manner? After all it's a very simple shape. no?

Maybe I'm just too new to Houdini to understand the need of this complex way of modeling...

 

So I'm wondering why you're not using the real nice, and way faster method provided by fredhopp with his AWESOME RASPBERRY ? foreach with a vop to peak(inflate) untill volume detected.

[Pancho]

The smoothing loop just takes care of the edges. No time lost there.

 

Asmuch as I like Fred's approach, it not fast either. I wonder what costs so much time in my approach. I like the way I try to handle the simulated modeling, but it's far from real time. But maybe H is capable of fast calculation times in this area. I once tried to copy a Maya flag in H. It was just sad. Maya's nCloth is much superior to H. I guess it might be between x5-x10 faster. So probably the way I chose is one of the really slow ones.

 

All other steps seem pretty straight forward and simple. I don't know how you would subtract the pearls from the jelly to avoid interpenetrations. With transparency and SSS this will make trouble sooner or later in the rendering pipeline.

 

Cheers

Tom

 

P.S.: After digging a bit into VDBs cookie sops are really one a second choice for booleans. It's better to go VDB and then remesh for a cleaner mesh with no wholes or errors.

[Pancho]

P.P.S.: As soon as the base shape changes you'll see the big advantage to do a lot of stuff procedurally.

[6ril]

 

I got no pb understanding the benefits of a procedural approach, don't worry. I just think there's a difference between procedural and over-complex.

[Pancho]

Anybody who can tell if the dynamic approach this way is correct or if I use a wrong set up. Would cloth be better compared to soild+fem?

[f1480187]

Still digging through an incredible fredhopp's example. With metaballs timings can be shortened to couple of seconds.

raspberry_metaballs.hipnc

[djiki]

... and little bit faster approach (still procedural). Just basic concept of berry-like structures. No details like hair etc.The idea to avoid slow calculation of overlapping intersections is based on cellular noise, precisely on Voronoi cells capable of generating natural structures. So instead to calculate for intersection, this approach uses such structures to generate intersection shapes, Latter, resampling and smoothing them with simple rounded CAP ending finalize blobs. Probably there is much better "blob reconstruction" technique by curves of V coordinate with skinning but  this is just basic concept and it cooks in a blink of an eye.  Scene link:

 

https://mega.nz/#!aNwCiRTA!njRC4gudikrNdSXRnMZ7axN2sLm_kpGdYbLBTNy2LqA

 

 

Edited March 17, 2016 by djiki

[Noobini]

this thread got me raspberry juice flowin'....so had a go in Max with Cloth...(+ MCG special sauce)

Now lemme try in Houdini...gotta learn Cloth/FEM first tho....

 

Edited December 7, 2017 by Noobini

[Noobini]

so here's the fruit of my labour in learning Houdini cloth...for the sake of speed, accuracy is compromised...but the gist is there...hey, just done Cloth for a day or 2...wadda you expect ?

 

vu_raspberry.hipnc