Spiky Hair + Smooth Hair -- (How might I effectively combine them when modeling?!)

astraldata

I have had lots of trouble trying to find a single, effective, method to modeling these low-poly hairstyles -- but none moreso than when I have to model something like Goku's hairstyle combined with a hairstyle similar to something such as this pink one:

What recommendation would you suggest (technique-wise) to model lots of hairstyles like those above? -- They have (mostly) organic shapes that tend to taper carefully to a point, but there are parts of the style that need to maintain consistency with the scalp too (and things such as the part in the pink hairstyle [or even in Trunk's hair!] throw me waaay off in trying to figure out how to approach them, topology-wise).

I'm a big fan of the interesting shapes that come from the various hairstyles in most of these types of animes -- but DBZ has got to be the clearest representation of organic sharp and smooth in a single hairstyle that I've seen, and it's made me come to the realization that I have no clue how to approach these esoteric shapes topology-wise without redoing lots of work by endlessly deleting edges and/or repositioning verts. I've considered doing a high-poly sculpt / boolean to get the basic shapes, then retopologizing over them, but I'm pretty sure that's not the standard workflow to make such clean topology.

There has GOT to be a better way to approach these sorts of shape transitions...

astraldata

I forgot to mention that I already understand the pink hairstyle is essentially a combination of Goku (lots of extrudes and vert merges) and Trunks' (a big half-sphere and vert pulls) hairstyles, but they can get to be a LOT more complicated than that:

And regardless of the complexity -- the inside of that labcoat girl's hair is a solid/clean mesh! -- But how? -- I easily get lost in the topology of it, so I can't imagine pushing/pulling verts one-by-one for this:

Are there any other approaches / techniques than either 1) making a high-poly version and quickly retopologizing that to a low-poly version for "sculpting" these kinds of shapes? -- or -- 2) endlessly dragging verts and edges around while constantly chopping up part of a sphere and then merging some cylinder verts together for these low-poly styles.  What do YOU do when you have to model something like this?

The two techniques already mentioned really seem like the only (freeform) ways to approach this problem for such complex (flowy) hairstyles like the last one, but I'm still not convinced that's true -- There's definitely gotta be some trick to this!

kwyjibo

This kind of stuff is just  a lot of standard modeling to be honest. I don't think there is a good way that doesn't involve a lot of finicky tweaking, deleting, merging. Depending on the hairstyle, I'd start by making some basic primitives that can be duplicated about to create the volume, and then turn that into a clean mesh. This might even involve some booleaning and cleanup. I don't think the meshes need perfectly clean though. You could easily get away with not merging connecting all the segments in goku's hair for instance. Also creating basic uvs before duplicating the segments would be a good idea. Also tools like maya's soft selection or blender's proportional editing are really useful for bending and tweaking the individual segments once they are in place to create the natural looking variation.

Neox

just break it down into individual shapes, this should make it easier for you to grasp it


once you have these, you can merge them. i mean each of these strands has about a dozen triangles. shouldn't be too complicated to model each of them.

astraldata

@kwyjibo & @Neox

Thanks for the input on this. I see what you guys are saying about piecing it together. That was probably the way I might have approached the labcoat girl above. However, I converted it to quads in Blender to make it more readable, and the more I study the model, the more I begin to wonder about the "airtight" nature of these hair meshes.

These definitely reek of some form of retopology to me.

As you guys have said, there's no reason to have tightened-up such a heavy mesh. This was what got me wondering about the method used and also why I am attempting to deconstruct its workflow. This mesh almost doesn't have any errant triangles, but it easily could have while getting equally passable results (and saving time!) There's a plethora of DBZ characters, so this would have been very important, but even Goku and Trunks up there were built the same "airtight" way.

The topology also makes too much sense for such a messy hairstyle in general... Even the (separate) CURL at the top of the hair is welded:

As you both have said -- there should have been no reason many chunks should have not been either left loose (or at least union-ed together, assuming they wanted airtight geo.) That fact alone leads me to believe that these were constructed in some other way -- perhaps with quads overlaying on top of (potentially?) messier (booleaned?) geometry underneath it all?

Has anyone else here modeled that way -- or is that a black art or something?

Neox

the amount of time you put into overthinking the approach, you could already have modelled a few of these hairstyles

yes it could have been retopoed, but why? create the strands, merge them together by hand or with boolean, done

kwyjibo

I've done a bunch of this kind of stuff using the methods I outlined above. It can be a bit finicky but with a bit of practice and a bit of planning it's not too bad. It is low poly after all so there's not that many vertices to deal with. Those meshes have not been retopo-ed. That I can guarantee you.

astraldata

Neox said:

the amount of time you put into overthinking the approach, you could already have modelled a few of these hairstyles

Haha! -- Fair enough!

This was mostly just analysis for fun. I tend to learn a lot about what's right and what's wrong with my approaches (speed-wise) by doing this sort of thing, so I really didn't feel like it was a bad thing to ask (just in case my chosen approach wasn't the best/only approach available!)

kwyjibo said:

I've done a bunch of this kind of stuff using the methods I outlined above. It can be a bit finicky but with a bit of practice and a bit of planning it's not too bad. It is low poly after all so there's not that many vertices to deal with. Those meshes have not been retopo-ed. That I can guarantee you.

Thanks for the words of wisdom on this.

Now I am wondering -- if this was assembled in parts (say, through boolean operations), is there an addon for blender (or something else) that can quickly clean up errant topology produced by stuff like that? It takes me FOREVER to clean up topology, even on moderately "simple" models, so that's why this particular model confounds me so much. Some input on this area would be much-appreciated.

After looking at this, there's *gotta* be a faster/easier method/tool/technique to merge verts/edges/faces quickly for the artist to have gone through the trouble of merging every single vert/edge together (while maintaining the exact shapes of everything!)

Are there any tools available that would assist in getting this kind of vertex/edge boolean/merged geometry cleaned and placed faster? -- I can easily model separate shapes, but it's the combining part that always trips me up -- especially when the topology is drastically different!

pior

You're still overthinking this. If your workflow doesn't allow you to effortlessly cut edges into shapes in order to merge them at will, then ... that just means that your priority should be in improving your basic modeling skills and smoothing out your toolset.

"This was mostly just analysis for fun."

This whole thread is pretty much pointless then. Why would you ask for help and advice about such thing if you don't intend to put in the effort yourself in the first place ?

For instance people here could record little helpful videos to show you how to treat these interconnections - but what good would that be if you are "just asking for fun" ? In that case anyone attempting to help you is literally wasting their time.

"I can easily model separate shapes, but it's the combining part that always trips me up"

Then I'd say try again, by having a stab at accurately recreating on of these DBFZ models.

astraldata

pior said:

"This was mostly just analysis for fun."

This whole thread is pretty much pointless then. Why would you ask for help and advice about such thing if you don't intend to put in the effort yourself in the first place ?

As far as I am aware, I never said I wasn't going to put in the effort to improve my basic modeling skills -- this thread was created for that exact reason. The fact that these were analyzed "for fun" doesn't necessarily imply that the *result* of said analysis is not *also* being taken seriously in addition to the "fun" to be had..

In short -- no, this thread was not in vain at all.

I always have fun finding different ways to approach my models. Analyzing other models gives me lots of data on how some models were put together compared to others. And asking others more experienced with this than me directly is an even better approach to learn in some cases. There's no sense in practicing doing things in the "wrong" way when there's a better way available.

By no means did I intend to say I am not taking anyone's advice here seriously -- I simply said I am having fun doing so.

That being said, I am looking for speed advice on general "low-poly" modeling (i.e. without retopo -- not low-poly "style"), particularly on how these larger (round or unpredictable) shapes are merged in a fast/easy way with many smaller shapes quickly. If you offer a video or any other advice on how to approach this, then don't worry -- it will definitely not be in vain.

Why else would I bother to make a topic to try to suss out the methodology behind these models if I weren't intending to apply what I learned from it??

pior

Just do it man. By the time it took you to write these couple thousand words you could have made progress already.

Also look at the title of this thread. It doesn't quite sound like an "analysis" to me - it's more like straight up asking for help. And the widely accepted philosophy around here is that the first step to asking for (and getting) help is to put on the big boy pants and putting in the work first.

Your move man.

astraldata

pior said:

Just do it man. By the time it took you to write these couple thousand words you could have made progress already.

Also look at the title of this thread. It doesn't quite sound like an "analysis" to me - it's more like straight up asking for help. And the widely accepted philosophy around here is that the first step to asking for (and getting) help is to put on the big boy pants and putting in the work first.

Your move man.

Thanks for the advice @pior (It's really good advice!), but the problem isn't that I can't do (or haven't already done) the work for what I'm asking -- it's that I want to find a better way than how I've already been doing it -- and for that, yes, I am asking for help.

The whole point of this thread was to ask for advice on a better method, tool, or workflow to approach the specific task of combining greatly-differing surface topologies at slightly higher poly densities than "low-poly" is usually for.

My current workflow speeds/methods are not to my satisfaction (i.e. bridging edges, merging verts, or boolean-ing complex parts together with no real surface similarities in the vanilla Blender way, adding lots of cleanup), and I see no real potential for genuine speed improvements outside of simply getting faster reflexes to shortcuts I assign -- and there's a limit to speed gains from that alone. I'm not getting any noticeable speed improvements out of continuing to blindly move/bridge/merge verts and edge-loops manually one-by-one via shortcuts anymore, so I really want to improve my workflow, methods, (or my toolset, as you mentioned) for this task. So any advice on that would be very helpful and much-appreciated.

I typically work at lower polycounts that aren't quite sparse enough to be considered "low-poly" (take the pink girl's hair for an example of density), however even these density levels are still too "low-poly" for retopology to be a worthwhile part of the workflow (just to make it "easy" to combine arbitrary 3D shapes), so I am trying to eliminate this step in my workflow. Any advice on tools / methods / workflows that could assist in quickly combining general lower-poly stuff (such as sharp to smooth stuff or sharp stuff to flat or round stuff) at these kinds of polygon densities (in Blender, mainly) would be amazing and, again, very much appreciated.

Alex_J

Too much thinking, man! 

The issue isn't how you are using the tools, it's how you are using the most important tool between your ears. You are treating this like some complex engineering problem that needs a novel solution. But it's super simple. 

There is a time for sitting down and using the brain, and there's a time for putting the thoughts aside and just working with your hands and eyes. Try the simpler approach first!

astraldata

That aside:

Regarding @Neox's comment -- I'm curious about this particular boolean method. Seems like there would be a quick way to merge the extra nearby topology to a clean edge-loop. Edge-loops I can work with. However, I also wonder if there's a quick way to reduce/increase the number of polys on an edge loop (or across an entire selection) in Blender... If this is possible, that may be a better approach. Maybe there's an addon for this?

Also, I have been looking into trying to find the right settings for the sculpt tool in Blender that would let one spot-check some retopology with that Dynamic mode in the sculpt brushes. This might make a good way to make better edge-loops where sharp-meets-flat on boolean-ed surfaces.

Has anyone tried something similar to this?

pior

"the problem isn't that I can't do (or haven't already done) the work for what I'm asking -- it's that I want to find a better waythan how I've already been doing it -- and for that, yes, I am asking for help."

Well, this is pretty much reaching the point of trolling ...

If you don't show what you've tried/how you've done it, no one can help you. The problem is not about finding some kind of new automagical method ; the problem is that you are probably either slow, or inefficient, or inaccurate, or all of the above when using basic techniques. Because basic techniques is all that is needed here - and that was confirmed to you right here in this thread by people who've actually done this exact kind of work.

I don't like using analogies but let's put it this way : you are pretty much going to see a doctor because your knee hurts and you can't walk, but you are telling the doc that no no no, he shouldn't look at your knee - instead he needs to give you a jetpack.