Home Adobe Substance

Triplanar seam hiding in Designer

polycounter lvl 3
Offline / Send Message
YannickStoot polycounter lvl 3
Hey guys,

I'm trying to solve a problem I have been encountering for quite a while now: Hiding your seams.

In this example I have a battery which has a concrete texture applied to it through triplanar mapping. Unfortunately there is a seam where the UV islands meet. 


Problem area:


My goal:
My plan is to blend these "perfect" edges with a normal UV mapped texturing method through a mask.


So, is there a way to create texture these edges without the UV-cuts?
(I don't care about blurring btw, it just needs to not be a sharp cut)

Replies

  • poopipe
    Options
    Offline / Send Message
    poopipe grand marshal polycounter
    Does your concrete texture tile? 
  • YannickStoot
    Options
    Offline / Send Message
    YannickStoot polycounter lvl 3
    Hey man! Yea the texture is tileable. The problem is on every seam though so I would assume the triplanar is not working 100% smooth? How do other people normally fix this in Designer?
  • poopipe
    Options
    Offline / Send Message
    poopipe grand marshal polycounter
    There's a blending parameter for the corners but it really ought to just work assuming your UVs aren't weird.


  • YannickStoot
    Options
    Offline / Send Message
    YannickStoot polycounter lvl 3
    Hah shit,... No idea to go from here then
  • poopipe
    Options
    Offline / Send Message
    poopipe grand marshal polycounter
    It's a bit hard to diagnose tbh.   I'm sure it can be fixed


    can we see the UVs  and your graph? 
  • YannickStoot
    Options
    Offline / Send Message
    YannickStoot polycounter lvl 3
    Hey man!

    Sorry for the super late reply. Have been incredibly busy.
    But yea you are right. I have not shown my problem clear enough for you to figure out what the problem is. I'll try again.

    What I am trying to create a node set-up that blends away the edges of UV seams. My thesis is that I can use regular mapped texture for the large areas of the UV space and then use a triplanar mapped texture for the borders.

    Tri-planar mapped set-up:


    Regular mapped set-up


    Complete set-up


    UVs



    As you can see, the set-up works pretty well but the tri-planar mapping is killing it with the shitty blend on the edges. Is there a way to remove these artifacts?

    If you need the work-files let me know. I'll send you a pm.
    And thanks for taking the time!
  • poopipe
    Options
    Offline / Send Message
    poopipe grand marshal polycounter

    Im thinking you could probably fix it by fudging the position information.  Send me the files if you like,  I'm curious about it now
  • YannickStoot
    Options
    Offline / Send Message
    YannickStoot polycounter lvl 3
    So I found the problem, sort-of. I just don't know what it is being caused by. What seems to be the problem is that the inputs that feed into the graphs do not contain the right data.

    This is my world space normal:


    Correct me if I am wrong but isn't this supposed to be a perfect gradient if the object is not moved after the bake?

    @poopipe
    Send you a pm.
  • poopipe
    Options
    Offline / Send Message
    poopipe grand marshal polycounter
    I've had a quick look in the file - ill look more thoroughly later but The model in that image has bad UVs 
  • YannickStoot
    Options
    Offline / Send Message
    YannickStoot polycounter lvl 3
    poopipe said:
    I've had a quick look in the file - ill look more thoroughly later but The model in that image has bad UVs 
    I've gone and simplified some things. Instead of the whole battery I am now only trying to materialize the wrapping. So the UVs have changed a bit:


    Question:
    Apart from a tiny bit of stretching near the edges, and straightened edges. What else could I improve upon?
  • poopipe
    Options
    Offline / Send Message
    poopipe grand marshal polycounter
    If we ignore the top surface then that bake looks OK. 
  • YannickStoot
    Options
    Offline / Send Message
    YannickStoot polycounter lvl 3
    poopipe said:
    If we ignore the top surface then that bake looks OK. 
    Haha: if we ignore the problem it looks OK. I have no idea why it renders out like this.
  • poopipe
    Options
    Offline / Send Message
    poopipe grand marshal polycounter
    The top surface looks like that because the uvs are offset and overlapping the sides
  • YannickStoot
    Options
    Offline / Send Message
    YannickStoot polycounter lvl 3
    poopipe said:
    The top surface looks like that because the uvs are offset and overlapping the sides
    Shit you are right. Still had them on the 1002 space and have the "is UDIM" to false. I will retry the whole set-up a second time. Seems I've been sloppy with my approach. Thanks for pointing that out!

    I'll try the full object/set-up a second time.
  • YannickStoot
    Options
    Offline / Send Message
    YannickStoot polycounter lvl 3
    I've fixed the bake output and put the output to 4k yet there still seems to be a problem with the tri-planar mapping.
    Is it possible that I need to add a more pixels through a UDIM set-up to wash away the hard edge of the UV?
    Or are there other ways to remove these hard edges?


  • poopipe
    Options
    Offline / Send Message
    poopipe grand marshal polycounter
    That's the bit I need to muck around with 
  • poopipe
    Options
    Offline / Send Message
    poopipe grand marshal polycounter
    more pixels isn't going to solve this I'm afraid.

    I've tried with your meshes, and I've tried with my own and I think you're hitting the fundamental limit of what triplanar mapping can do -
    The blending is inherently crap when the surface is at or around 45 degrees from the projection axes, which happens at your nice edge bevel (and 4 times around the cylinder) .
    and.. 
    The tiling is inherently crap if the corner you're looking doesn't line up with the tile rate of the texture . 


    I'm not sure where you're aiming to get to with this stuff -  if it's just for this asset,  you can solve the problem by unwrapping it in a straight line (including the horizontal lip ), not using triplanar mapping , accepting the stretching and rejoicing in the fact there's no seams

    If you wanted something that's more of a general case solution I think you'd need to start looking at building a multiplanar mapping node - 
    there's a 3d planar projection node in 2018.1 that could probably be leveraged.  It'd still have weak points halfway between each projection axis but they'd be in narrower areas

    I'm curious now.. 


  • YannickStoot
    Options
    Offline / Send Message
    YannickStoot polycounter lvl 3
    I'm mainly testing out a new node I want to make to simplify my texturing process. In the case of materials without large scale details it does not matter that the corners blend, nobody will notice. However if there is a massive seam then people will. If I can have the edges of the UV's blend nicely I can UV map much faster and only be precise when I need to.

    The perfect solution is of course 3D materials but that is only use able in some situations and not all materials can be made with 3D nodes. (not even talking about the fact that it is only released for about a month)

    I understand what you mean by multi planar but I feel this might result into the same result planar as I have now since a single planar mapping node also has the artifacts.
  • YannickStoot
    Options
    Offline / Send Message
    YannickStoot polycounter lvl 3
    I've just cleaned one area and left one area not cleaned up. Seems like you are right. This actually is the best this node can do. I do wonder what causes this since it it should theoretically be feasible through this node. I really wonder how one would solve this though. 


Sign In or Register to comment.