Faster and Easier - Studio d-Force or Blender Cloth Sim?
zombiewhacker
Posts: 683
Since I don't have a dedicated graphics card, I haven't been able to really dabble with either Studio's d-force or Blender's cloth sim. I've been wondering, though, if I bought a d-forceable clothing item, would I be able to cloth drape it in Blender, and how would the results differ? Also is either method faster and easier?
I only know that Blender doesn't require an nVidia card, whereas apparently for d-force an nVidia card is a must.
Comments
Why purchase a dforce garment? When there are so many free ones you could try. Yes, they might not be the quality you want or even the garment you want. But they would be an inexpensive place to start. I would recommend watching a few youtube videos on animating cloth in blender first. If you have not done so. Keeping in mind. That Daz Studio props are made for Daz Studio. And may or may not need adjustments to make them work in blender the way you want. Mainly welding seams together or seperating some. I would also suggest trying some none dforce garments. That way you can get a comparison on usablity. And there are a lot of free ones also. Once you have a method down. Then concider making purchases. I would even suggest watchng videos on how to make cloth for Daz studio. Seeing how it is done may give you some additonal information to work with. Start with something one piece. With no buttons/decorations attached. Or remove the buttons/decorations. A flag is a common place to start. And there are a lot of examples available online. Isn't there a flag model with the deforce plug-in? Just remember you are learning at least two things, 1. How to animate cloth in blender, 2. How to convert daz studio cloth for use in blender. Which one you do first is up to you.
As for which is easier or what the differences are. The one you work with the most will always be easer. For me dforce is easy until you stare adding wind forces. Then it is out of control. While adding wind, gravity and colisons in blender are more workable for me. Getting to that point is harder for me. Your milage will vary. Two different processes are always going to be different. In the way they are executed and final results. Learning one can sometimes help with the other.
Why would cloth sims require a GPU? They're math calculations; it has nothing to do with video.
Blender's cloth sim has more options and is more robust, at the cost of increased complexity. Since cloth sims depend on the geometry of the mesh, they will deform exactly the same. The difference is in the algorithms used to control the bending and tension. I haven't done any 1:1 comparisons, but I can't recall Blender ever exploding the way dForce does. As I said, it also allows you much more control to fine-tune things, as well as being fully integrated with Blender's other simulation and particle systems, of which there is no equivalent in Daz Studio.
Computer video is also maths calculations. GPUs are optimised towards the KIND of maths calculations used in these simulations. nVidia has a software that makes these maths calculations easy to use on their GPUs. DAZ uses that software for cloth simulations, Blender doesn't.
THAT'S why!
And now to correct the poster you were responding to: No, an nVidia card is NOT a must. It just makes things faster.
GPUs are optimized the store textures and render them really quickly. That's not at all the same as typical physics simulations, which can be easily computed on the CPU.
A GPU is also hardware, not software.
You are twenty-five years behind the times then. Since then transformations in 3D space (or as it is commonly knows, 3d acceleration) has been the driving factor.
Acceleration of vertex rotations and translations in bulk are exactly what GPUs are designed for and incidentally are what physics simulations use.
And? Hardware is accessed by software and you are still twenty-five years too late when you expect to deal with the hardware directly in a user-level software, instead through an API.
Here, have a look at what GPUs are capable of doing:
You may also like this.
https://bitbucket.org/Diffeomorphic/import_daz/issues/410/better-dforce
I appreciate all the feedback. Has anyone here done Blender cloth sims with no dedicated GPU? (I'm currently on a i3 budget laptop.) Wondering if I took a DAZ character and clothing attire into Blender and did a drape what the process time might be.
I don't think Blender uses the GPU for physics at all, unless they changed it recently.
I have a Ryzen 5 quad-core, 3.2GHz.
I tested a subdivided plane with about 3,000 verts draping over a sphere. With default settings, it runs twelve frames per second. With self-collision on, it goes down to 10 FPS.
There is no clear answer for how long it will take as each simulation is different. How dense is your mesh, what's the topology look like, what quality settings are you using, are there collissions, self collisions, how dense are the colliding objects, etc. On an stock Threadripper 1900x, I aim for 2 seconds max to calculate each frame. If it takes longer then I re-evaluate my settings.
With the dforce default settings, that are intended to be the reference for all the dforce outfits, blender on a ryzen 2200G takes about twice the time of dforce on a gtx 1060. On par of quality with the same scene. So if you get a good cpu (the 2200 isn't) then blender can compete fine with dforce. That said I myself would love to see the blender simulation at least with opencl, I believe going with the cpu is a little "prehistoric" nowadays.
https://bitbucket.org/Diffeomorphic/import_daz/issues/410/better-dforce
I've definitely exploded blender simulations - though freezes are the more common failure
But I agree overall, to answer op neither has best in class speed, blender is more robust and you can do way more once you know what you're doing, I would say Dforce is much easier to set up simple stuff nd has things ready made for you
blender also has the absolute best option to optimize for low specs and simulate complex things > create a proxy mesh > simulate that > use a surface follow modifier to drive the more complex mesh. you can simulate anything with simple geomety buuut it does require being able to do some modelling
wanna see something nifty you definitely definitely cant do in DS?
does anyone know how to embed gifs? haha!
You don't even need to model anything. Just slap a decimate modifier on a copy of the original and crank it up until it's low-poly.
It is best to use a clean mesh to drive the HD mesh anyway. In the example by JCade I guess she used a subdivided plane that's a good mesh to drive the cloth simulation over the HD chains.
It is also worth noting that driving the HD mesh with a proxy is not always a good thing. Especially for clothing. Specifically, we need the HD cloth geometry to simulate wrinkles. If we use a low poly proxy the simulation is faster but we lose details.
lower poly is good if you the default res is taking ages to render but yes it will obviously sacrifice some accuracy - I dont always use it super lower res all over sometimes it will even be about the same density, but just without mesh details like seams, hems, pockets on a mesh not designed for simulation, thin straps that stretch weirdly, etc. I've even kept chunks of an original mesh that sim well and keep them but then edit other parts of the mesh that are problematic
and yeah in that example I used a simple subdivided plane I even lined it up with the more complex geometry. For my final test every repitition of the complex mesh fit over 4 polygons
So... anyone have a favorite tutorial or documentation on general workflow of applying cloth physics to Daz clothes that have been imported into Blender? :)
and again ..
https://bitbucket.org/Diffeomorphic/import_daz/issues/410/better-dforce