I thougt that psyfi was hands on the multires issues.
Hello,
thanks for your work and also, thanks for cummunicating with the users.
My opinion is that combination of grease pencil + weight paint would be optimal.
There is, how noted allready, huge problem with weight paint performance. But I think that this should be taken as a separate problem and the feature should not be hacked together with mask painting from sculpt mode . It simply doesn’t belong there, or does it? Just an opinion…
I actually think masks, weights should be merged into one feature, and used together, like what I propse here:
http://wiki.blender.org/index.php/Face/Edge_groups_proposal
and the engine for these should be the same, so performance can be optimized for all of them…
I like the original proposal.
I’m pretty sure he’s doing this approach because he has a lot of experience in laplacian algorithms (like the one last year that allows for some level of posting for dyntopo sculpts).
Also, this thread is already in danger of becoming nothing more than noise (like usual in BA), because there’s a few who are upset that he’s not working on exactly what they think should be upmost priority. Psy-Fi said he was eying a possible multires overhaul on his todo, so give it a rest and let the guy with more experience in sculpting code tackle it.
Also, my opinion, yes I do think it should be at automated as possible like others said, the general flow of the shape should also exert tighter control over the laplacian fields than what the paper suggests if possible (ie. minimizing the unnecessary wandering of edge loops).
@Ace Dragon, I am a noise maker now? Really?
From my point of view, blender becomes the more suitable application for those who are interested in development only.
Not for artists?
There isn’t any serious plan on real developing. Every developer is working independently.
Little or none collaboration, not much effort on unification of the tools.
The real problem in blender, every possible workaround is facing a dead end, sooner or later.
Then comes Andrew Price and the UI blender issues. I never followed such approaches on this matter.
Noise? We may have a lot of it soon.
In topic now. It looks like a hack when asking to use mask brushes… However it is the only brush that performs well on high density meshes. I agree though, remeshing doesn’t belong in sculpting mode.
We’re talking about remeshers and retopology tools, right? I missed something? Such tools are needed to work on high density meshes in most of the cases.
So, we will have a remesher soon, that’s great. But I can’t answer on questions like “should we use vertex or weight painting”?
Maybe in two years, maybe never, who knows.
Scultp, retopo and modeling are linked, depends of the workflow of the user.
A fully automatic remesher can be usefull, but not for everyone.
A remesher is not only for multirez, so we need control.
I use zremesh for parts of meshes and some times for all meshes, I dont use multrirez after because I need a good low poly and add it a sub surface modifier.
After if this Gsoc is only for a fully automatic remesher, ok ! No more demands !
So, for the topic, the simplest way will be the masking bursh I think.
It sounds like you are acting based on preconceived prejudices. People are just expressing their opinions, chill down. Blaming others for not being able to propose the ideas the way you prefer/want wont take this thread anywhere.
Actually fully automatic remesher can be useful for everyone, if it is done right. It sounds like you use Zbrush. Zmesher is pretty good in most cases, at least it takes care of %50 of the problem (ie general topology coverage in quad terms) in most cases which means that one has to deal with the other %50 of the issue in the worst cases in average. So it definitely saves time.
There are cases where automated/semi automated solution is not desirable, however I tend to think that these cases are minority compared to the cases where one can take advantage of full/semi automated retopo solution.
Sometimes Zmeshes does a better job than average retopo job at higher end of the low res topology (ie > 3000+polygons). it wont beat hand made super low res geometry most likely.
I use Zremesh, but mostly blender for hand retopo, So I need more controls than autoremesh.
Because autoremesh is limited, it’s why I ask for some other controls in this gsoc.
Every people don’t have the same attemps to a remesher, A fully auto remesher will be usefull for everyone, but I think, it’s controls peoples needs more than automesh.
You don’t sculpt on an autoremesh, it’s a waste of time, so you retopo yourself for some parts.
Autoremesh is usefull for certain kind of modeling, not characters, for caracters it’s hand retopo only.
And if I retopo, it’s only for characters and if we sculpt it’s mostly for characters too.
After, like I said, if this GSOC is only on full auto remesher, no need to talk about controls or other ideas.
it wont beat hand made super low res geometry most likely.
Never. Agreed. But, we need a spontaneous tool, to convert dyntopo into a multires procedure.
Real re topology comes after.
From my experience: I developed a rather fast method to manually retopo dyntopo sculpting. For multires needs only.
A low poly base as possible. This won’t be the final result, for sure.
From my experience, testing either 3dcoat or zbrush auto retopology: They work well when asking for rather dense bases. This is not compatible with the multires modifier.
On a face, a portrait: I can manually build a 1000-1500 faces base. Multires will work fine up to millions (surprisingly it will go up to 25 millions, zbrush limit = 8-16 M faces per mesh). However, using 3dcoat or zbrush remeshers, basemesh will cost 4-10 k faces. Multires can’t beat zbrush then, sometimes it won’t let you work on more than 4M faces. (it depends)
Actually we do not need a dyntopo to multires atm. We first need a remesher, then fix multires, then fix/improve/replace shrinkwrap( so it works as good as Zbrush projection) Once those are in place, autotopo to multires happens by itself.
Although I understand the purpose of using weight paint tools to define the remesher, it feels too time-consuming and I don’t like the idea of going into a different edit mode just to prep a model for remeshing. Ideally I think a solution utilising the Grease Pencil tool to divide the model into easily mesh able segments is the best solution, as it can be used in any mode and would generally be quicker to use. Also, if what other users are saying is true, that performance in Weight Paint mode is poor with a high resolution mesh, then such an approach would be useless for high-polygon meshes, as well as Edit Mode which from experience has crappy high-res performance.
The modifier has less value as a tool if it can’t operate on high-polygon meshes, and i’d recommend Grease Pencil as it doesnt suffer the same performance hits as other tools have, but it depends on what kinds of models you’re designing the modifier for I guess 
I think it should be made possible to have a fully workable mesh that can be animated through a retopology procedure, isn’t one of the points of multires anyway to make a displacement map to apply to what would be the base formed by that retopo?
I think that’s something to consider because a lot of people might mainly use multires sculpting for smaller details that’s harder to do with dyntopo, as the latter is already very good at macro-detailing since you wouldn’t run into the issue of running out of fine geometry to play with.
Also, if we were to use empties to define positive and negative areas, there should be a way to just have them in two different groups and allow for the use of the different empty shapes to define different shapes of influence (a circle-shaped empty around the base of a limb to denote the entirety of it as a negative pole).
I thought about the Empty and grease based approaches, and they might work, but proximity calculations might be just as slow as weight proximity modifier on high res meshes. Since quad remesher is a modifier it will suffer from the same evaluation problems as any other so the min/max should be optimally stored somewhere, weight map or a custom data.
Actually we do not need a dyntopo to multires atm. We first need a remesher, then fix multires, then fix/improve/replace shrinkwrap( so it works as good as Zbrush projection) Once those are in place, autotopo to multires happens by itself.
Agreed.
However, we first need a vertex based painting system, capable to work on high density meshes.
BTW, shrinkwrap works similarly to zbrush project. Once again, all the issues start from the base mesh. A very low poly mesh as base won’t perform well under shrinkwrapping or under zbrush project.
There is a very serious issue regarding the base mesh displacement under multires.
i’m confused.
I thought the purpose of a remesher was to make evenly distributed quads, to get rid of stretching when sculpting. Like dynamesh in zbrush.
This is what the current remesh modifier is supposed to do, but instead creates ugly diamonds. The remesh modifier isn’t a “retopology” tool by any means (as in, sexy loops for animation or deformation), it is what it is, a remesher.
someone jump in and correct me if i’m wrong.
@ng
I think we’re going for something better than this. A nice flow of these quads as possible. If some triangles or n gons happen here and there, it isn’t really important. Expecting something ready for animation is a little hilarious.
I doubt if it will work well when asking for 500 faces assets as well. Such algorithms aren’t capable for the task.
Yes shrinkwrap is similar to ZProject, that is not the issue. The issue is that Shrinkwrap is not very usable in most cases because of performance and the cleanup needed afterwards(whereas ZProject does not create as much messy projections) . ZProjects performs very very well on low or highres or super highres meshes and it is pretty fast at highres meshes. You really do not want to try Shrinkwrap with highres meshes.
Also Zproject is much superior because you cna just maskout and project unmasked areas quickly with highres meshes. You cannot do that with Shrinkwrap, unless you want to torture yourself by going into edit mode and trying to select and create weightmaps with highres meshes. Well the otherway is to convert sculpt masks to weights, then maybe yeah ok, one can try to use as a knockoff ZProject.
In most cases Zproject is used transferring high details to highres, high to mid, mid to high etc. And you can use it for lowres meshes too. Shrinkwrap is just “ok” with lowres meshes, if i do not need dynamic projection I just go into edit mode face snap enabled and do it manually.
I don’t want to sound too negative, but I really don’t see this project useful.
Having a good remesher is a great thing,but in my experience even the best tools are not “magic”,so an older approach I doubt it can be useful for real case.
Anyway,as others have allready said,the best approach should be scalable.
No edit mode,no vertex paint(at least until vertex paint is so slow).
The remesher don’t need to work on really huge meshes(as high frequency details will go on texture/displacement) but at least on medium details meshes.
An auto retopology tool, being able to create very low poly meshes (mostly quads) is our only chance to use blender multires sculpting.
Let me explain further.
Supposing we have sculpted a nice full of details multires model.
What you gonna do with it? Bake normal or displacement maps?
You can try BI or cycles baking. Multires method.
BI is not able to bake tangent maps from level 0 (for a very good reason*)
Cycles is able to bake from level 0 only (too bad *)
- multires modifier distorts the base badly (if apply to base), or doesn’t displace it at all (not apply base)
This base can’t be used as reference for multires baking, it goes worse, you can’t duplicate and use this base as bake to activel.
So. your only real option is the bake to active mode, consider though, you have to perform a second retopology on top of the multires sculpting. Two manual retopologies are too much, don’t you think so?
Ace Dragon, I’m pretty sure you can follow my thoughts and understand why me, and Brecht actually, have the idea to trash multires and try a new one, re written from scratch.
All these are in topic.
What’s the use of an auto retopology tool in blender?
I own zbrush, it’s easy to test such workarounds, found not much use though.
Blender should be a tool for art, not a tool to demonstrate developing skills.
IMO this is the obvious difference between blender and zbrush. Pixologic always develops tools for artistic needs only.
Please don’t take me wrong. The development of this auto retopology tool was always a dream, a wish.
It is simply useless in blender the same way Zremesher is in blender.