Interesting idea. Mixing different shapes on the same topology.
Would ‘node group operators’ be limited to Edit mode?
Are there plans for something like mesh generating node group operators we could call from object mode?
If so …
Base meshes could be hidden behind a GN operator asset. The user does access the GN operator asset, tweaks parameters as he likes it. He does not need to know that several meshes with identical topology are used and mixed by the GN operator.
In the generated mesh, there won’t be a GN modifier left.
I has got some feedback about lowpoly multires basemesh concept in our local group which could be considered as negative.
So I decided to make a deeper practical research and use it for actual sculpting by myself to check the concept.
I uncovered the tablet, spent a couple of hours and sketched up a male head which could technically be considered as “planar”. Here is the result:
Overall, I found this concept pretty useful using it for multires sculpting especially from references when global shape should follow exact references and the most of local details has to be sketched manually, and I liked the ability to tweak the base in edit mode to fix some initial shape issues like this which I discovered after sculpting (the original head basemesh will require those changes as well):
Also I found pretty funny nominal backward shapekey compatibility. A quick and dirty proof of concept test (sculpt in multires ones and have all the variety of shapes for free at easily tweakable base), looks brutal and I like it:
@JulienKaspar could such a planar head sculpt considered as useful for this bundle?
According to the feedback that I got, the resulted anatomy of this sculpt is not perfect (like an ear is too low and the nose is too much straight) but I think it is nice for a sketch and it could be improved in further I guess.
Here is the file for check (Bundle Planar head shape test by 1D.blend)
Well, it was a pretty intense week, I think we progressed well. I think I will need to run some tests of this topology in order to enhance it a bit to make it more versatile. Have a nice weekend!)
According to workflow design, the initial stage - sketching - is the most flexible one, because any other following stage assumes following technical restrictions.
Sketching is time to unleash creativity, so it is reasonable to have as much flexible sketching base as it is possible.
The main benefit of a multires over dyntopo is preserving the initial topology, which allow to flexibly modify the resulted sculpt by homeomorphism.
I run some test of a basemesh head and found several issues which I tried to fix.
I added additional loop under jaw (to support possible bodypositive shapes) and to the ear and also unified the shape of an ear across shapekeys to achieve a versatile topology transition solution.
Unfortunately, those changes was pretty deep and everything has to be done nearly from scratch, including pretty precise sculpting which I especially don’t like to do mostly because of…
So I did it. The left one is old original and the right one is resculpt:
The planar head sculpt has been chosen on purpose to check how exactly details are transferred during transformation.
The result is pretty fine, for example ears are transfered much better, so all those rework was worth it.
I also added Realistic_Kid, Ork and Suzanne shapekeys to test extrema.
Anyway, sculpting in multires across the multiverse is an interesting experience)
If I hadn’t left the sculpt years ago, I would have definitely been glad to use this base head mesh.
I also added Toon_female blockout and refined/revisited existing ones, for example added ears and nose to toon blockouts and foot fingers to everyone.
Great work!
The only thing missing I think are objects for the eyes. Those would be good to include. I’m thinking of simple spheres and half spheres like in this example video.
Some edge loop definition like for finger nails can also be removed and left simple instead.
I also though about that) Large eyelids are a parts that are better be shaped separately.
Unfortunately, when I tried I faced with eyes depth problem, and when I tried to solve it I’ve got such kind of a custom solution which looks suspicious.
Different blockouts solves this place differently.
The eyelids appear to be parallel. From the side, their gap is rectangular. Please check side view, eye corners usually are v-shaped.
An artist could fix eye corners, but it is an extra-step he has to do if he starts fresh.
It is technically possible to reconstruct the original shape of an eyelids, but I don’t know how much it fits the basemesh requirements. It looks very predefined)
Nobody:…
Absolutely nobody: …
1D: Here is a toon female model unsubd iteration!
In order to get a better source for basemesh parts (like nose and palms), better understand the topology flow/junctions and also investigate how this model may look like before subd I decided to make unsubd retopo of a toon female model. The result is 0.92 subd, pretty close to original:
Much more helpful, imo. It’s great for learning and specially useful for those with low spec computers in places with fewer financial means, for example.
I prefer to think of it as of a cleaner topology guide.
The model is very clean, so the body technically can be usubdivided even more… The ability of such a reverse engineering looks tempting.
I’d be up for replacing the base mesh topology for something lower res. The deciding factor for me is multires performance. There’s an annoying factor right now that if the base mesh is too low/high poly, the performance takes a hit. I’ll look into it if there’s a negative impact with them.
I think of unsubd of it as of easily deployable base, it is usually useful to have them because of a lower amount of control points, which is good for curvature setup - so you can shape it faster, apply subd (if needed) and start sculpting.
Also it is a nice topology guideline.
At the moment it is just interesting topology research.
I think the best we can do is to iterate what iterates at the moment to have the ability to test how much it may be useful in practice.
For example, realistic female model could not be iterated that easily.
I was wondering if it is possible to unsubd stylized female head even more, to reach a base which could provide the topology similar to original after double subd, and here is the result:
It is possible to reach a base which provides topology similar to original with subdivision, but for sure it doesnot provide same details shape because of smoothing. The resulted model looks very analytical)
I also reached the lowest form of a body and transferred it across the other shapes:
The result is not that bad as it actually could be, being as lowpoly as possible, I think I reached the limit.
Palms was the most challenging part, it was quite challenging to redirect loops in the way which would allow to connect pretty dense fingers with lowpoly wrist without lots of topological messup.
LODs are shapekeyable, but I decided to not to go shapekeys this time because of a possible topological research changes and improvements.
It will be difficult to optimize them even further.
Here is the file:
Bundle Skeleton+Blockouts+Head+LP_v07.blend
Thanks @1D_Inc !
I’ll have a look at these. If they work much better I’m also fine to transfer the detailed shapes from the subd levels onto a Multires. That way we have the best of both
Recently I found base mesh with featureless face and shape keys that let me morph between skinny and fat on all body parts, and it has become only base mesh I use for actual sculpting. No matter what body type I have in mind I can get something closer and start working, it has been relevatory.
So I am in favour of shape keys, yes. I understand it might not be obvious for starters, but that shouldn’t be a deciding factor I think. Not everything has to be for absolute beginners, professional users are using Blender and those assets too, and besides without those sorts of stuff to teach you and expose you to new features it’s harder to advance in 3D.
Can you provide a screenshot of such a base? It is interesting how it looks like)
I am not sure if it is worthful to provide transferred model in a bundle - the goal of LP base is to provide a versatile shape for sculpting, but it will be definitely interesting to see the results of such a test anyway!
Deeper tests brings a better picture of a possibilities.
For example, I expect some problems with lips and ears during transfer, and it is interesting how it will work in practice.
I also found a way to improve LP a bit.
The triangle on the foot and the pentagon on the chest are connected with the same loop, so they both could be eliminated by connecting. This will also provide additional geometry for the hips, which is sort of needed and is better to have.
Speaking of a further topological improvements, being inspired by the solution provided by @dan2 female model, I also decided to rework back topology of a body LP, in order to get a better control over resulting subd shape.
I am getting more and more busy recently (for example, UNESCO approved our historical reconstruction project which is sort of a win, and now we have to iterate it further), but LP simplicity which we reached before allow to analyse and work with topology faster, which compensates busyness.
It makes it affordable to analyse its topology in background.
Love the movement towards as low poly as possible. I think this will be great for people who want to modify the joints for animation tests and modelling accessories.
