Zbrush - Is it possible to pose after this?
node
Hi there guys and girls! I'm kinda new to 3D modelling. I've been modelling/animating cubic (minecraft like) models in Maya for the past 1-2 years, but never a smoothed out high poly models. So I took the very hard decision to get me some Zbrush and this is my 1st try:
http://imgur.com/9M1zhA0
So, as you can notice, my model has 3.75mil points and there is no active slider for Sdiv, which means that I can't go lower that this anymore. I think I used Dynamesh the wrong way. Anyway, the idea was to sculpt this model in a T-pose rig it, and make a custom pose (why not even an animation) with a base below it, water and it's name and stuff...like a collectible statue, big Pokémon fan. But after I rigged it with Zpheres and tried to pose it...it was incredible slow process + the mesh deformed very strangely.
So my question is....is it possible at THIS moment to pose this model without ruining it's detail or not? I'm okay if I have to wait after every pose move I make, I'll pose it for 2 hours if I have to...I just want the mesh to be smooth, not deformed. In zbrush / maya...doesn't matter, I just want to pose it like I had in mind...cuz I feel like I worked 4 days just to hear "you f*****d up...you'll have to start all over"
By the way...that's what I'm talking about:
http://imgur.com/eEE4yXU
I tried to Dynamesh it on a lower resolution, I lost nearly all my details, but i was able to pose it super fast...but even then the mesh deformed and bugged kinda awkward and I noticed it was places that are RIGHT below some very big "bones"/parts of the Zphere rig...which I think means that some parts of my Zphere rig are TOO FAT and near the mesh itself right? Maybe they need to be thin and buried very deep inside the mesh?
That's it I think...please I will greatly appreciate if someone can help me with this cuz I really really wanna see my model posed and "alive".
P.S. will send beer to the person who helps me
http://imgur.com/9M1zhA0
So, as you can notice, my model has 3.75mil points and there is no active slider for Sdiv, which means that I can't go lower that this anymore. I think I used Dynamesh the wrong way. Anyway, the idea was to sculpt this model in a T-pose rig it, and make a custom pose (why not even an animation) with a base below it, water and it's name and stuff...like a collectible statue, big Pokémon fan. But after I rigged it with Zpheres and tried to pose it...it was incredible slow process + the mesh deformed very strangely.
So my question is....is it possible at THIS moment to pose this model without ruining it's detail or not? I'm okay if I have to wait after every pose move I make, I'll pose it for 2 hours if I have to...I just want the mesh to be smooth, not deformed. In zbrush / maya...doesn't matter, I just want to pose it like I had in mind...cuz I feel like I worked 4 days just to hear "you f*****d up...you'll have to start all over"
By the way...that's what I'm talking about:
http://imgur.com/eEE4yXU
I tried to Dynamesh it on a lower resolution, I lost nearly all my details, but i was able to pose it super fast...but even then the mesh deformed and bugged kinda awkward and I noticed it was places that are RIGHT below some very big "bones"/parts of the Zphere rig...which I think means that some parts of my Zphere rig are TOO FAT and near the mesh itself right? Maybe they need to be thin and buried very deep inside the mesh?
That's it I think...please I will greatly appreciate if someone can help me with this cuz I really really wanna see my model posed and "alive".
P.S. will send beer to the person who helps me
Replies
You are currently running into a common roadblock experienced by many people jumping straight into Zbrush. But your experience with Maya for block-like models will undoubtedly be useful going forward, so that's great. There are a few different ways for you to go from there, I'll try to sum things up. This is all assuming that your starting point is this one dynameshed model.
- - - - -
Option A : Staying in Zbrush and attempting to directly pose this dynamesh model using transpose, zsphere structures, and additional sculpting. This is basically what you are doing already - this is doable, but as you probably noticed it is very frustrating, fully destructive, and does not lend itself to experimentation. It's good enough if the goal is to produce one single pose (for a statue print for instance) but definitely not sustainable if you want to create multiple poses - and impossible to animate.
Option B : Reworking the structure of the model using various Zbrush reduction techniques (decimation master, zremesher, maybe some manual retopo) to get it to be as light as possible. This will give you a medium density model, looking almost exactly like a sculpt but much lighter. This means that the posing tools inside Zbrush will behave a bit better ; and, if you reduce it low enough, you will also be able to assign it to a proper skeleton in Maya for relatively non-destructive posing. Some parts will definitely glitch out and will need manual smoothing back in Zbrush. You will be able to make multiple poses rather easily, but Maya will struggle.
This was posed this way : https://cdnb.artstation.com/p/assets/images/images/004/342/191/large/pior-oberson-oberson-polovov-elf-high-001-jpg.jpg?1482692316
Option C : Completely rebuilding the model using traditional subdivision techniques, just like it is done for cartoony animated CG movies and colorful game cinematics (Smash, Pokemon). That means that you don't necessarily need to go as polished in Zbrush to begin with - you simply establish the rough shapes, then take that into maya and rebuild all the parts with clean polygon surfacing. You can then pose this lowres mesh in Maya, and this will be animation friendly too since subdivision will be performed as an effect on top of the model. This will give you the best visual quality.
I was checking the models done by Pony fans and they do them that way - meaning that the models are light and easy to share, but can be very nicely rendered once posed and subdivided. http://i.imgur.com/U2xFpKD.jpg
Ironically it has been possible to create models like this way before the Zbrush days - it just took longer to get the forms right back then because things like smooth, polish and grab brush didn't exist. But the end results were clean, subdivision-friendly models that are easy to pose and animate, and smoothed out at render time. And when it comes to the clean Nintendo art style, a well-built game model can also be converted into a subdivision-friendly model rather easily. Like some of the shots here :
Option D : Creating a continuous "game-style" lowpoly model (either manually or using zremesher), creating full UVs for it and transferring the texture data of your Zbrush sculpt to it (colors of course, but also normalmaps if you want to catch all the subtle surface qualities). That means that you end up with a realtime model that can be rendered in a game like environment but you will be limited to texture resolution and this is overall a rather complex process (much more complex than what it first seems). Like this : https://cdnb.artstation.com/p/assets/images/images/005/427/955/large/pior-oberson-female-mannequin-cubebrush-010.jpg?1490928193
- - - - -
And ... that's it I guess ! Bottom line is, creating a detailed Zbrush sculpt really is the easy part. You have various options from there, but you'll have to experiment quite a bit to get to a smooth workflow. I highly recommend you to rely on C, that is to say taking a few steps back and rebuilding/resurfacing your model using a traditional subdivision workflow. Using creases for sharp edges might be an option too.
Good luck !
A is not an option I guess for me at least
B is most likely how I will proceed bcuz I have older versions of my model that have every element on a different subtool which means I can reduce only the body to as light as possible, fix a few parts that will go wrong and merge at the end. Every blue part that you see on the model (except tummy) is around 30 to 50k points max, which I think is light enough...I hope, and some of them can get lower too. Only the body has 2mil and the legs 200k. Question IS...do I have to just merge all the parts at the end by clicking MergeDown OR should I merge them for real using dynamesh...where they actually merge meshes and become one mesh? I HOPE that's not necessary. Cuz things like it's blue face mask need to stay high polly for some close face renders.
Something else I clearly never knew up to this point is that...a model can be lowpoly like Sdiv lvl 1 and for the render only to go highpoly like Sdiv 4-5, is that right? Sorry for being so noob, but I never really needed to use this for my cubic models. I just made em lowpoly for my game and for rendering purposes I was adding divisions and smoothing out.
So if that is true...we kinda go into option C that you are recommending. If I rebuild completely the model in lowpoly I can somehow export it to Maya for the posing and render that will jump into Highpoly only for the render. Will Maya understand/can control those Sdiv levels that we use in Zbrush (I mean the slider with levels 1-2-3-4)? How those Sdiv will translate into Maya? There are some things I don't understand and how they work, but I'll have to watch tutorials and learn. I'm sure many people are doing this. Sculpt in Zbrush and then to Maya for pose/animation so...lots of tutorials I hope
Option D is something I have never encountered yet. UVs, texturing and stuff aaand you said it's a hard option so...I'll just stay away from it.
About C, I saw the Pony screenshot and noticed how all the parts of those models are kinda...rounded and smoothed the same way...am I explaining this right
Actually this is a good moment to ask...how many polys combined (or points if we are talking in Zbrush) should I have, to pose it without having issues like really slow or destructive mesh? What's the standard here? I know it's kinda hardware dependant, but I have a high end PC with OCed CPU and SLI so I guess I can go higher without worrying too much.
I will try to do Option B and see how it goes while I wait for a response. Option C is the best, but I've spend too much time on this model to start all over, I just wanna do something else. So if B can be pulled off...I'll do B and after I'm done posing it I'll start a new project with everything you said in mind ^^
- - - - -
"Something else I clearly never knew up to this point is that...a model can be lowpoly like Sdiv lvl 1 and for the render only to go highpoly like Sdiv 4-5, is that right? Sorry for being so noob, but I never really needed to use this for my cubic models. I just made em lowpoly for my game and for rendering purposes I was adding divisions and smoothing out."
"Will Maya understand/can control those Sdiv levels that we use in Zbrush (I mean the slider with levels 1-2-3-4)?"
- - - - -
There are 3 ways to do this, lets call them C1, C2 and C3.
C1 : simply smoothing/subdividing a lowpoly model, without any further care. This will indeed gets you a overall round and melted look, which I suppose is what you noticed on parts of the pony model.
C2 : subdividing a polygon model made especially for that purpose. That means controlling hard edges using tight edge loops along the edge of a detail that needs to be sharp, or, using edge creasing values (from smooth to fully sharp) to indicate sharp edges. Maya has that. (Maybe the Pony armor parts originally had such edges creasing info and the data might have gotten lost during various model conversions, not sure).
A good example : http://wiki.polycount.com/w/images/thumb/9/98/Subdiv_gunstock-racer445.jpg/512px-Subdiv_gunstock-racer445.jpg
Note that both C1 and C2 do not require any information stored/sculpted at higher levels, since all the sharpness information will come from either edge loops positioning or edge creasing.
C3 : creating a detailed source model, creating a subdivision-friendly lowpoly base with UVs (just like in option D), and capturing the details of the high using a displacement map, which deforms the surface of the model at the time of subdivision, at render time. This is how highly detailed CG characters are done in fantasy and scifi movies/cinematics. Example : http://www.yuri3d.com/uncategorized/hello-world/
Now there is yet another option, which is very elegant and doable by default in Max and with a little bit of extra work can also be done in Blender and Maya. Lets call it C4 :
C4 : Basically 3DSMax has the option to subdivide a model accordingly to hard edges placed on a lowpoly mesh. This means that one could build a very light model, place hard edges (aka "smoothing groups") where needed, apply a subdivision modifier to keep these hard edges and smooth the rest, and apply another subdivision modifier on top to gently smooth out everything. This is often referred to as "double smoothing".
You could replicate that in Maya and Blender by simply building a regular lowpoly with soft and hard edges until it looks good, and then select all these hard edges and mark them as sharp creases for subdivision.
- - - - -
All that said, B is indeed the fastest option for you at this time, even though it will likely by awkward and painful to pose. You could start thinking of ways to achieve C on a next model, or using small parts of this current model as test cases. Edge creasing + edge loop control is definitely the key here ; look closely at behind the scenes model shots from Pixar/Disney movies, they are great sources of inspiration.
One important thing to keep in mind is that there are really two aspect of developing a sculpted character : the artistic aspect of the sculpt itself (using clay like animation often do in the earlier stages, or Zbrush as a replacement of it), and the technical aspect of it (building a version of the same character that can ultimately be posed and animated). In a production environment these two tasks can very well be taken care of by different people altogether.
Good luck !!
"I'm sure many people are doing this. Sculpt in Zbrush and then to Maya for pose/animation so...lots of tutorials I hope
As a matter of fact ... very few people are doing this. Most Zbrush-centric artists either pose/decimate/render their sculpts directly (that is to say, option A again) ; or, if they are interested in the game pipeline (which is pretty much D) they build a dedicated game res model, bake albedo/AO/normals, and texture the model like one would texture a game model.
What you are describing (using Maya to pose a low and render it as a high) is the movie pipeline, which relies on displacement maps for details. Very few Zbrush artists know how to do that, mostly because generating accurate displacement maps requires a bit of technical know how (models need to be at the proper scale, the displacement map needs to be saved in a very specific high bit depth format, generating the displacement map is better done in a dedicated app outside of Zbrush, and so on). This is why I would recommend you to eventually learn how to use creasing and edge loops - because models made that way are trivial to subdivide and render.
I hope this makes sense
And I'm not sure I understood, do I have to use the MergeDown option in Zbrush for all my subtools or I should merge them down for real into one mesh using the dynamesh method ? Cuz MergeDown options makes them look like they are together and if your brush happens to catch em together, it deforms them KINDA together...but really they are not together. And dynamesh merges them into one mesh for real.
I think MergeDown will do the job cuz it kinda combines...groups them together.
Oh and another thing about the poly/point count. How much is acceptable for a good posing? Are my 30-40k subtools too high polly? Should 500k all combined mesh be good for the posing or it's still too high and heavy to work with?
P.S. where to send the beer to
Edge creasing is for the Maya side of things - in Maya you can simply select an edge and give it a value between 0 and 1. Creasing in Zbrush is another thing altogether, very destructive and very likely impossible to export out. There are ways to convert Maya creases and OBJ hard edges into Zbrush creases, but probably not the other way around. In other words Maya creases are made especially for clean rendering of carefully built models, whereas Zbrush creases are mostly just a tool to quickly get a sharp edge and keep on sculpting.
"Merging Down" subtools in Zbrush basically brings them all into one mesh, but it will not weld anything in the traditional sense (unlike dynamesh, which indeed will melt everything together as you noticed). A "merged subtool" still has "polygroups" though, meaning that you can isolate parts at will and even split them out again later. You can also use the MoveTopological brush to only affect a part that's under your brush without deforming the one right next to it (great to lover an eyebrow without affecting an eyeball).
As for mesh density ... it's up to you to decide what is responsive enough as far as both Zbrush posing and Maya skin weights painting/posing are concerned. The heavier the model the less responsive it will be, so that all depends on your own tolerance levels. For what it's worth you can use the file size of exported OBJs/ZTLs as a rough indicator of how heavy a model is. If it is in the hundreds of megs it will probably be near impossible to skin weight in Maya ; if it is around a few dozens of megs it should be okay-ish.
No need for beer or anything at all
http://imgur.com/4n8KgVJ
http://imgur.com/SnufBml
It smooths okay (too bubbly,but I can work with that...try to learn to crease edges and stuff) only problem is the way it bends. You see how it is in the middle of it? How the middle part is a lot lower than the left and right part of the finger. On the back, its okay...but not in front. You can see how thick my Zphere rig is...so the main question is - is my rig the problem? Is it either too fat/big or too close to the mesh? Maybe both? Do I need a much slimmer and well positioned rig...will this fix the problem? If yes...I'm doing a new one, posing, fixing, puting a nice base with water and stuff, render...and upload somewherw with HUGE thanks to Pior ^^
Oh and another thing...I tried to find a tutorial for this, but failed. If you happen to know...how to put the rig points so the wings bend like...a banana...not 90°
Also once you have that, you can subdivide this newly Zremeshed version and even capture the fine details of the original sculpt using Project All. You basically end up with a reconstructed model featuring good-ish lowpoly topology (from Zremesher), subdivision levels, and information from your original sculpt transferred to the higher levels.
I can't give advice on zsphere rigs since I don't use that.
https://sketchfab.com/models/a06381f112d64504a5817d6406333675
It's not finished, I will add backgrounds, moon, color/material everything...but just wanted to share how I did this with masks...and reposition the pivot point. I think I am a great masochist, but also achieved something in a unique way