Hello everyone. I am not currently that proficient to know how to go about this setup, and if someone knows if there is a tutorial that go through this setup, most importantly explaining it thoroughly, please write.
Anyway, if someone could help me with some guidelines on how to approach this, much appreciated. Here is the picture that will hopefully explain in a visual way what type of scene i want to make.

I modeled 15 different flowers. I need these flowers to be floating through the space, mainly falling with the gravity.
Even though there are 15 of them currently (will make some more as i go), i want to have a bunch of them, for which i guess the only way is to instance them, in some efficient way, with some pop's. As there would be a bunch of them falling, some would probably have to collide with each other, which to me implies the need for rbd also? Since they are flowers, they should not be stiff, so atleast the petals would need to be swinging, not much, just little is enough, and for that it would be the vellum setup?

Efficient, beacause i am currently working on a laptop with Intel(R) Core(TM) i7-4720HQ CPU @ 2.60GHz 2.59 GHz and 12 RAM )

What would be the route to do this?

I say go for vellum. If possible lower the amount of polygons on the mesh and at the end use a "vellum post process SOP" to get the smoother/high res versions by using the subdivision option inside that node.

There is also another way which sometimes might not work. It is to turn the meshes into just lines (like an L-system tree) by putting a convert line after the mesh and then using a fuse to combine it together. Then using a wire solver to simulate them. After the sim use a "Point Deform SOP" to get the mesh back from the wire simulation.

you'd most likely go with proxy vellum sim and then deform hi res as suggested

instead of wire, which wouldn't work well for petal collisions nor for pointdeform, I'd go with
either simplified petals, which you can then use to deform more hi res ones
or
simplified bud and sim as tet softbody to make and then deform hires bud

it depends on the amount of deformation you need and whether you need separate petals to behave completely idependently

tamte
you'd most likely go with proxy vellum sim and then deform hi res as suggested

Hello Tomas, Berk!

I was doing some research on vellum in order to understand how it works, and i oriented my workflow as suggested.
The problem is that the pointdeform is really not performing as it should be, or my setup is wrong. It transforms only one geometry and not all that i scattered.
Notice the squashed mesh thats located in the center of the correctly transformed mesh, it looks messy.
Note that the flowers are made of parts that i simply merged.

Can't really say what the issue is without really getting into the scene.

Maybe try to bypass some nodes and see if deform works, like bypassing randomization by weight or change the scattering options. Just to find the cause of the problem.

Another idea that I have might be to use Boolean and fuse to combine the flowers together. (After merging the parts of the flowers put a Boolean SOP and plug into both inputs of the boolean then choose union, then add a Fuse SOP)

What if you completely remove point deform, just use a low res geo for vellum then make them smoother/high res with vellum post process?

That's all I could think of, hopefully Tomas will have some better ideas. Will get back to you if I can think of something else.

Hello Berk! I just saw your reply.

Well currently, how i see it, is that since i have scattered points and i use those as instances for vellum objects, when i try to point deform the setup probably gets confused because im picking both the scatter stream for the 2nd input of pointdeform and points from the DOP as the third input. Plus i have another copytopoints for the sake of hi res objects, that go in 1st input.

Maybe i need a rest position but then again im scattering every some frame...in the OdForce i found a thread where there is pretty much the same setup, but honestly i don't understand it quite well. There he used sop solver after the scatter and simply merged the last frame with the first input, and plus he uses a pointdeformcapture from the Ql library...i dont want to complicate it with the long post, so I'll just paste here the link to a post from there.

https://forums.odforce.net/topic/45056-vellum-simulate-with-proxies/ [forums.odforce.net]

Difference is that i'm having multiple different objects, and he only deals with one (a car).

Honestly i haven't tried how it would look if i go all the way with the low res and postprocess it after. Im going to try that approach nevertheless.

Ps. If the working file would help, i could attach it then..

I am checking the odforce post.

Waiting for the results of low res and post process thing from you.

In the meantime if I can think of anything else will get back to you.

MilanB
Ps. If the working file would help, i could attach it then..

it would definitely help you as you would probably get it back fixed
it can be simplified version not necessary your production file

but essentially you either have hi-res geo(s) pointdeformed in a loop for each of the corresponding proxies
this is useful if you have a few geo's duplicated multiple times so like 3 source models and then emitted into sim resultng in 100
so for each of those 100 proxies you pointdeform corresponding hi-res

or you have hi-res geos in the same count 1-1 with proxies and have the same piece attribute
and then pointdeform using piece attribute from rest hi-res without for loop

in both cases it's beneficial to have hi-res in rest pose and also having some attribute for rest for proxies, so that you can get aligned rest pose for pointdeform after sim