A question about retopology, high poly characters and low poly ones...

polycounter lvl 2
Offline / Send Message
oraeles77 polycounter lvl 2
A question about retopology, high poly characters and low poly ones...


I have been learning with Topogun to retopograhpy high poly models. all of my stuff is quite simple static objects.

I have been looking at models on Artstation of game characters.

the models seem to be made up first of a naked human character model, then a layer of clothes are put on top using Marvelous Desginer, then extra accessories, such as shoes are put on top perhaps using Zbrush etc.

When all of this is retopographised into a low poly game ready character, what happens (what should be done) regarding cavity with in the model, for example the space between the clothes and the character's body, etc?

Its something which is bugging me.

Replies

  • Mark Dygert
    Generally it gets combined into as few contiguous meshes as possible. So a belt for example would be modeled into the pants with just enough detail to help it hold it's shape. There wouldn't be any hidden faces that belong to the belt A tie, vest and coat would be combined.

    If it helps, think of your retop geometry as a shell that you're putting around your high poly. 

    There are exceptions to that. If the character is going to have some level of customization (different shoes, ties or bow ties, t-shirts or coats bla bla bla) it can affect what you combine and how it all flows together. 
  • Alex Javor
    Offline / Send Message
    Alex Javor hero character
    I am a beginner like yourself, so don't take this as the word of god, but I'm working on a modular character system for a Unity game at the moment and can share some ideas: 

    First thing I made was a base model, and this is a dude in his underwear. This model would only be used if the player removes all clothing items, but another reason to make this model is that it is my mannequin for putting clothes onto in MD. So it gets textured and rigged and then it is ready to do its job in MD. 

    Next I make the different pieces of clothing in MD and/or Zbrush. Once I have the game model of the clothes ready, I delete any of that base body mesh that exist underneath the clothes. It cannot be seen so there is no reason for it, and it will help me avoid any clipping issues during animation. Then I go through with the texturing, rigging, and so on.  

    I am not sure if this is how pro's might do it or not. I got the basic idea of it from an archived thread on this website. Today I am just beginning to test switching out the different meshes in engine, so if there is any major problems I'll update this so I'm not putting out bad info. But, if you question is just basically, "what do you do with the mesh underneath clothing/exterior meshes?" I think like Mark Dygert said you only need the outer shell that will be seen.


  • poopipe
    Offline / Send Message
    poopipe ngon master

    The in game mesh really only needs to define silhouette and support animation - if something doesn't contribute to those it's surplus to requirements. Exactly what's required will depend on your project but it's a good rule of thumb. 

    There are exceptions in the real world - customisable characters are a great example-  but it'll usually be down to needing flexibility in the system or a matter of logistics (is there time to build all the variations you'd need to make it properly efficient or can we afford to cheat a bit to get more content in game?) 
  • oraeles77
    Offline / Send Message
    oraeles77 polycounter lvl 2
    hey, thanks for the replies, but I still can't find the answer to what Im looking for,

    for example, these renders...

    https://www.artstation.com/artwork/o5RkL


    this isnt a game ready character, but a render, but my question is still the same and this particular area is the one gray area of my understanding of 3d character design.


    1) make or obtain a nude character, either from Make Human or Zbrush/Blender by hand or import it from somewhere...

    2)  rig the high poly model (if you want to use the high poly for rendering etc)

    3) put clothes on it in in Marvelous Designer.

    4) add the accessories like shoes, bag hanging on the shoulder etc.

    5) make a low poly retopolgy (I use Topogun but theres loads of others to chose from), rig that, unwrap it, burn the textures etc and export it to your game engine.



    now, Im sure there is something missing between 4 and 5, cause all of the amazing character models on artstation I see using Marvelous Designer, I cant work out how they'd retologise it,


    another example...

    https://cdnb.artstation.com/p/assets/images/images/003/022/933/large/madina-chionidi-back-withharness.jpg?1468608410

    if you look at the straps on the torso, they are above the fabric, in some places they have cavity, how do you retogographise stuff like that?


    sorry if I come across as a bit dumb!





  • oglu
    Offline / Send Message
    oglu polycounter
    there is a good example for retopo complex cloth on artstation... but nothing for beginners...

    https://www.artstation.com/artwork/zZOOL
  • oglu
  • Kanni3d
    Offline / Send Message
    Kanni3d polycounter lvl 4
    They're all separate elements, and retopologized accordingly. Anything that would never be seen can get culled to save on texture space/vert count in engines (an example would be hands with gloves... the hand geometry would be deleted entirely since it isn't seen). Floating elements get their own geometry for better fidelity and bakes (generally with higher quality games) which would simply floating on top of the underlying geometry (another example would be like a pair of jeans with a belt. For lower end games, you could retopologize these two elements as one, or for more fidelity, you could retopo the belt separately, delete its backfaces, and have it float there on top of the pair of jeans.)

    Hope that helps! I think that's what your question comes down to. :)
  • oraeles77
    Offline / Send Message
    oraeles77 polycounter lvl 2
    Kanni3d said:
    They're all separate elements, and retopologized accordingly. Anything that would never be seen can get culled to save on texture space/vert count in engines (an example would be hands with gloves... the hand geometry would be deleted entirely since it isn't seen). Floating elements get their own geometry for better fidelity and bakes (generally with higher quality games) which would simply floating on top of the underlying geometry (another example would be like a pair of jeans with a belt. For lower end games, you could retopologize these two elements as one, or for more fidelity, you could retopo the belt separately, delete its backfaces, and have it float there on top of the pair of jeans.)

    Hope that helps! I think that's what your question comes down to. :)


    it sort of helps, of obviously if your character has gloves you can delete the hands etc, but I mean what is the work flow for that sort of retopgraphing regard belts etc? sorry if its a weird question, but as I've never been formally taught 3d design there are lots of big holes, and recentlly after experimenting with putting complicated meshes into Topogun and having  a complete mess come out the other end, I can see that maybe I shouldnt bother with attempting lower poly clothes/characters for now...
  • poopipe
    Offline / Send Message
    poopipe ngon master
    The ideal is a single contiguous mesh with lumpy bits to define silhouette.

    If a belt or any other element never needs to animate separately or go away just build it into the body mesh. 
  • Aydhe
    Offline / Send Message
    Aydhe polygon
    Ok, so i'll ask question too.
    When you're working on highpoly, do you think its good idea to skip body details if you know that your character won't be customizable and covered deep in clothes. 

    Also, completely abandoning body during retopology stage as well, since it won't be exposed.
  • goekbenjamin
    @Aydhe i can only speak for myself but:
    when I know that the part of the body will not be visible, I don't see why this part should be detailed
    ----------------------

    As I read, a "single contiguous" mesh is preferable.
    At the moment I have a mesh (pants to be precise)


    As you see there are buttons on the pants, when I now retopo that as a single mesh,
    and the rigger/animator will bend the leg, and one of the buttons is in the near of that bend area, the "baked button" will be stretched I guess?

    Will that make sense to make a separate "buttons" mesh?
    Or should I place the buttons like that I know the bend area will not be near the buttons?
  • Aydhe
    Offline / Send Message
    Aydhe polygon
    It depends on how you skin the leg. if you have button in the area of stretch then it is quite likely that you will end up with a stretch in texture.
  • CheeseOnToast
    Offline / Send Message
    CheeseOnToast polycounter lvl 14
    Try not to place small, rigid details in a high deformation area if possible, as you said. A separate button might create a different set of problems such as leaving an AO spot behind as the knee bends, or clipping through the parent mesh. This is usually much worse on very low poly models. On denser models you're much more likely to have vertices on the underlying mesh close to the floating detail. Copying  the nearest vertex weight from the underlying mesh to the floating detail usually works quite well.

    Ultimately, game art always has some compromises. You'll often see "rigid" armour pieces deform a little, even in high-end games. Try to design the model with the end goal in mind, to avoid the worst of the problems.
  • kanga
    Offline / Send Message
    kanga interpolator
    CheeseOnToast said:

    ...

    Ultimately, game art always has some compromises. You'll often see "rigid" armour pieces deform a little, even in high-end games. Try to design the model with the end goal in mind, to avoid the worst of the problems.
    That is a really good point.
    If you have a character with a costume it helps to break the figure down Into what  substance painter calls sets. You can use the same strategy with any baking package. The advantage with substance is you can have your separate meshes in one document if they are named correctly. What determines the content of your sets/meshes is that there is no adjacent geometry that will result in backing artifacts. The two main methods are explode baking and sets. Explode baking is where you move parts far enough from each other so they dont  interfere with the process. The advantage with explode baking is you can keep the entire character as one mesh. The disadvantage is that after baking you will have to move all the parts back into place. The problem with this is matching the parts back isnt exact. You could use snapping to fix that I guess.

    The second method is sets. You break up your model into separate parts or meshes. Its important that the parts overlap enough so that when the character is animated there are no holes when the parts are moved: like shoulders and a vest. To fix that make sure there is enough overlap. It would be a waste of texture space and poly resolution to overlap parts too much, but bad for appearance if the model shows see through spaces when in motion. Below is a character Im busy with in zBrush. The costume has loads of overlap.


    On the right of the image all the parts are broken down into subtools, I made a copy and decemated them separately then  merged them into groups for retopology in Blender. I then deleted the lower subdivisions of the original tools and merged them into identical groups, named the groups with the substance naming convention and exported them as an FBX. I then sent the decimated groups to blender for retopology and then UVs for the retopologised low poly groups. I then imported an FBX of the finished Blender models into substance painter for baking and texturing. Below are the baked sets in painter. You can see the separate groups and the mesh overlaps.


    The result is a clean bake. It doesnt matter if the model is in parts with whole uv sheets for each part because rigging in 3dsmax for example allows placing a copy of the skin modifier in the rig stack. So one skeleton can be skinned to different meshes and animate the meshes with the custom skinning of each part.

    A variation on this is to break the high and low poly character into parts for baking after doing the UVs. That way when you assemble the parts again there will be no uv overlap. Thats what I did with this cartoon like character in Qixel below.



    Hope that helps.
  • goekbenjamin
    thanks alot for your effort and insights!
    When exploding your mesh for baking, don't you "give up" on the ambient-occlusion?

  • kanga
    Offline / Send Message
    kanga interpolator
    @goekbenjamin, yes on certain parts. There are really so many maps and options generated that Ive never found it crucial. However there may be a solution to that too. I just havent found it yet. Perhaps someone will chime in.
  • CheeseOnToast
    Offline / Send Message
    CheeseOnToast polycounter lvl 14
    If you use Marmoset for baking, you can make an "AO dummy", i.e. a highpoly which has no equivalent lowpoly piece. It will be ignored when baking normal maps etc, but included in the AO bake. I'll try to post an example later on when I have more time, if that description makes no sense.
  • CheeseOnToast
    Offline / Send Message
    CheeseOnToast polycounter lvl 14
    Sorry this took so long. Here's a quick example image. I am making use of baking by name in Marmoset. Each of the clothing elements has a _high and _low component. The skin part only has a _high component. It will only be included in the AO bake (with Marmoset's "ignore groups" option ticked on), and will be ignored in all other baked maps. I've never tried it, but I think it should be replicable in the Substance suit as well.

  • oraeles77
    Offline / Send Message
    oraeles77 polycounter lvl 2
    woah, people are still commenting on this post?!?
Sign In or Register to comment.