Early Impressions on Volumes


#1

I’d be interested in getting the community’s impressions of the new Volumes. I’ll log a few (very early) impressions.

-As you build and make changes to a volume there can be a delay while the volume is calculated or updated. At the bottom left of the screen you see the status… e.g. calculating %. Upon completion of the processing the viewport is brisk. I’ve yet to see how things go if adding five or ten volumes in a scene.

  • The default resolution is 10. I briefly tried .5 resolution and found that very unusable. It was beyond sluggish and sometimes the viewport stopped displaying the object entirely. I’m not surprised by this. The technology has it’s limits. But I do suspect I may want to change the default resolution from 10 to 4 or so…and then dial it still lower for renders. There are ‘range’ controls (accessed by twirling the control down), that impacts the resolution. You have discreet options for interior, exterior, spline and particles.

The example above had a resolution of 1.

  • The resolution settings impact more than just the quality. They change the shape of the final object…and the res also impacts the effect of smoothing and other operations. There may be times where we might prefer lower resolution in combination with a smoother. Very high resolutions can create a meta-ball type look.

-If you wanted a real crisp hard edge…I’m not sure how you’d get it with Volumes. The issue is that a smoother is almost a requirement to eliminate crunchy edges…but smoothing will round edges.

-The reality is that Volumes create a lot of geometry. This inevitably slows render times. You can decimate the resolution with the Volume Mesher’s ‘Adaptive’ setting but there are times there is too much quality loss with that approach.

-I did one experiment exporting a meshed Volume and using the free Instant Meshes app to assess quality. With this scene the results weren’t very good. Perhaps I need to learn more about using Instant Meshes.


#2

I have an overall positive impression, the way is integrated with fields and the rest of C4D is amazing. For example employing particles and splines and volume them.
that said this is what think are limitations and what i think it needs improved.

Before all that ICM you should put the adaptative to 1%, that is by definition that value that should not change the geometry - i have found what i think is a bug in an example i posted in another post but otherwise it respects that.
To geta crips edge you need to give a low resolution, like 1 to the voxels. It is the voxels the ultimately define the edge, are you sure your object in image is with 1?

  • It can be pretty heavy to work with to get the desired quality, for example i would not advise to build a full room with volume objects with current hardware we have.

  • the adaptative works and preserves quads but the form of those quads are not friendly to further improve the model, i don’t know if integrating something like instant meshes would be a valuable option but i think it is important that further improvements can be made on the model after the volume operation, for instance if one has a model that needs some boolean operations but the overall form is better defined by other modeling techniques. As it is now, implies that the booleans can only be made at the end of modeling, but there are many objects that need booleans at every stage.

-it needs to have a choice to handle bevel better , someone posted here the skirt option of Modo,. Currently it is ok to make a generic object is not the same when a client says : i want this precise object in the scene, then that bevel curvature of that specific object needs to be addressed specifically.

PS: i still not have tested with sculpt and other options with volume modeling.


#3

I dabbled with the volume builder /mesher for 1 day. I tried modeling a simple hard surface prop and found that you need a very high resolution to achieve smooth surfaces with sharp edges.
i doubt that VDB is appropriate for precision modeling. At best you can use it to create a base model that you can then retopo.
What’s clearly missing is a crease option.


#4

I am finding it is very much like the xparticles ovdb mesher with some additional tools to refine. Still playing though.


#5

I wasn’t aiming for hard edges in the example I posted. I’m still early on with this.

Yes…agree.

Yes. I think Volumes will rock for organic and abstract modeling and be useful maybe 1/3 of the time for more pricise technical items. I have to think in the future if they could run it off the GPU that you could perhaps get down perhaps to 1/2 unit resolution.


#6

don’t want to come across as overly dramatic but at this point i’d like to share that r20 - volumes came to the rescue just in time.

client: has a fair up and coming, needs some renders. stills done with maxwell and animations rendered with octane.
delievers cad-data, everything’s good, job works smoothly, beend working together for years now.
so… wednesday client comes up with a final job: needs a maxwell-render (still) for a press-release, a close-up shot of an invention the client wants to present on the fair.
needs to go to the printing machines on monday morning.

well… client delievers cad-data on thursday, i do my stuff and am ready to fire up the 2-day maxwell-rendering, when client feedbacks the preview and mentions that a detail of his product needs a change:
a small part of the product (2cm4cm2cm) had been changed by the engineers, which of course needs to be incorporated into the scene before the rendering process can begin.
it’s some kind of a blocking element that blocks another element of the product from being opened arbitrarily.

that small part consists of intricate details, eg. there’s your serial numbers, font-sizes would be around 0.02cm0.04cm0.02cm per letter. and other details like the recycling logo etc…
so, that small part needs to be shown in a deformed condition, showing a situation in which the “product” is in an opened condition.
by pushing that small part with a thumb, one bends it to a state where the blocking functionality is overridden and the “gates” of the product are being opened.

the nature of cad-data being triangulated messes would have left me in a situation in which i’d have had to rebuild that thing by hand in order to have a somewhat even polygonflow, so eventually bending/deforming-actions could have taken place.

i am not even sure, if i would have been able to do that. in that short amount of time left, considering the long render times, maxwell is featuring.

i opened r20, voxelized that small part with a voxel size of 0.001 cm, put a smooth layer (lap flow) on top of it, gave the mesher a go with adaptiveness of 0.05% and had a meshed volume-reconstruction of that part that - given the camera distance (though close-up shot, still far away enough) - was not really distinguishable from the original cad-data.
the rest was your usual bending/deforming-voodoo.

the whole thing took me 7 (!!!) minutes.
converted that structure to a polygon-object giving me about 10mio polies.

sent preview to client, got my “go” and fired up the rendering (due to the fact, that most extrenal renderers don’t have their r20-versions ready yet, rendering-process needed r19).

i was so impressed by this, that i even did not think about business: no extra-money to be charged, no nothing.

sorry for lengthy post.


#7

only played around with it for a couple of hours until now, so i might change my mind about certain things in this post.

it’s a bit slower than i expected, so i won’t use it as much in animation as i’d like to. not just because of the speed, also because of the dense meshes.

i think mesut stated a very good use case above, product vis is probably the area i’m going to use it most in the future. i often get cad files from clients where some parts are missing, like connectors, knobs and potis, stuff like that. i usually don’t need to get that close to those elements and most of the times those products won’t be animated, and if so they are quite heavy in poly count anyways, since they’re imported from cad, so a few thousand extra polys don’t make that much of a difference.

also i think it will be used everywhere metaballs were used before, for instance organic medical illustration/animation stuff. and it will be heavily used for 3D printing i guess. and of course for abstract organic mograph stuff.

i saw a bunch of facebook and twitter posts where people said stuff like “now i don’t need to improve my modelling skills anymore, i will do everything with volumes from now on.” i think that’s a bit naive. try to populate a complex scene with vdb models, even if made editable you will hit the wall pretty quick and the time you saved in modelling will be swallowed by vp slowdowns and additional render time. we will always have to think about the trade offs, personally i’m just going to use vdbs if it really makes sense and think twice about it. nevertheless i think it’s a great feature, and i guess we will see some improvements to it in future versions.


#8

I agree with all the above. VDB is a nice addition but it is far from the modeling revolution that I had hoped for. I wish there were more controls over the mesh and a good decimation/retopo system (I tried the poly reduction generator).


#9

I’ll put my 2 cents in as someone who used this feature during beta phase for my private work.
I do a lot of technical modelling for 3D printing, machine parts etc… Those parts need to be precise to 1/10mm and are usually not larger than 150x150x150mm
Open VDB allows me to stay procedural for much longer, often right until i export to STL for printing, than the mix of Poly by Poly, Parametric and SDS modelling i did before.
Since i know the ultimate precision i need i can model using a much more coarse resolution than i use for exporting, this makes working in the editor quite fast.
The layering allows me to round those parts that print and look better rounded, while keeping sharp edges for those parts that need them.
All in all my productivity increased quite a bit.
Just as an example, the mold for this Silicone tire is fully parametric, the parts in the back as well.


#10

Nice feedback Mesut

Srek, i am not following the process, how can you separate the model resolution from the export resolution?


#11

I agree with Bullit and Srek - you are going to have to use the adaptive resolution, and I think the Volumes feature is ideal for 3D printed projects… I do believe the following will improve the feature in the future:

Add feature for selections: material, polygon, and edges for the final result from the Volume Mesher when there are more than one polygon objects listed in Volume Builder.
I want control over the volume builder edges between objects and for it to retain material selections between different objects - its a bit limiting at the moment to have the output become a single mass without any material, polygon, and edge selections after the mesher creates the polygon version.


#12

That is an excellent point Continuumx. I guess for now we have to split the objects.


#13

yeah but for those selections we’d first need to have clean edgeflow, which we don’t. but that’s just how the vdb tech works in my understanding. i think if we had edge preservation with bevelling options we’d look at a completely different technology, more similar to mesh fusion in modo.

about splitting the objects or making manual selections as a workaround: won’t work, since there’s no way of controlling the edges. even if you work with projection methods you will get jaggies, that will probably only work for far away objects and/or pretty high voxel resolution. so to me as of now vdb is only suitable for single material objects.


#14

I’m simply using a much lower resolution fo rmodeling since i know that in the end i get the precision i need when increasing it for export. I don’t even need the volume mesher for modeling.

As for texturing and selections, OpenVDB does not come with tools for this directly afaik. It will require quite a bit of work and time to find a good solution here.


#15

I think i understand now, so you do all in voxels and only in final pass so to speak you build the volume mesher and export.


#16

For me, for my line of work. They look like a great way to get high-res quick conceptual pieces into Zbrush for further refinement.
I skipped R19, because it has nothing I want, but I"ll be upgrading this week for this feature alone.


#17

In the end i am still left wondering why something like hardops, boxcutter, meshmachine, decal machine for Blender do not exist for C4D.


#18

Like ProRender, Nodes, and other newer features, MAXON is likely to refine it for R21. Their mode seems to be to get the core functionality implemented and stable and then refine in subsequent releases. For my part I haven’t used it near enough to be qualified to comment on how good it is but for what I intend to use it for (more organic / abstract work) I think it will be pretty good. For some of that super-detailed stuff we’ve seen examples of recently from ZBrush and other apps, I don’t think this is built for that sort of thing yet.


#19

It was more of a lamentation of C4D polygonal modeling since in the end does not matter what technique we use to achieve the desired end result.

I agree this seems to be more for a less complex type of object. Integrating it well with sculpting, improving the volume builder adaptive resolution and way to control edges bevel i would say should be future development.