Home Technical Talk

What is causing / how can i avoid this baking error?

I thought this issue was caused by me decimating too heavily, but it seems even after redoing the high poly & decimating less the issue remained. I am getting black triangles on certain random faces on the high poly (i fixed it in some areas by manually merging these) and also this long artifact.

My mesh doesn't have a UV seam nor sharp edge here so it doesn't make sense to me, the bake goes fine in Substance but I cannot paint skew like in marmoset so i need to fix it here.

I'd really appreciate some info on avoiding / fixing this, cheers



Replies

  • Eric Chadwick
    Overlapping UVs? What does the low-poly model look like with the normal map on it shown full-bright (as if it was an emissive map)?
  • feint
    It's not that, I packed in Rizom and the blue circled areas are the only mirrored UV islands on this mesh (i did find some inner faces that i didn't notice so that was good to delete, but it didn't fix the issue)

    I think the screenshot below is showing what you mentioned?

    y m


  • poopipe
    Offline / Send Message
    poopipe grand marshal polycounter
    is there a seam there ?  what to the UVs look like?
  • dimwalker
    Offline / Send Message
    dimwalker polycounter lvl 16
    Geo verts in high/low mesh have splits there maybe?
  • Thanez
    Offline / Send Message
    Thanez interpolator
    Looks like missed rays to me. Have a looksiepie at your baking setup and check to see that the HP is fully within the confines of the cage.
  • feint
    Ok so, I decided to open my bake from marmoset and look at this area on the bake and it looks like the high poly must have faces that overlap and therefore cause an issue?

    Although I had already re-done the decimation to a lesser degree and the bake looked fine in marmoset for the most part, looking at the normal map in photoshop, it's really bad. I think this is ultimately still due to decimation. I was having issues when initially dynameshing my scope that there wasn't really enough resolution even at 4096, which makes it hard to completely negate the blockiness that you can see on the high poly in the second screenshot below.

    Time to be even less harsh i suppose. I'll give this a shot tomorrow and post about the results, appreciate the ideas on what to check for though guys.



  • Fabi_G
    Offline / Send Message
    Fabi_G high dynamic range
    Hi! Looking at the last screenshot, I'm wondering why the highpoly is not smooth-shaded? To ensure it's not related to something else, you could isolate the part showing the issue, see if it persists. That normal map looks to me like there's is overlapping geometry or UVs going on? Ideally share meshes (isolated elements showing the issue should be enough), so people can reproduce the issue on their end, cuts out the guessing.
  • ZacD
    Online / Send Message
    ZacD ngon master
    It almost looks like it's trying to bake a mesh that's the lowpoly and high poly overlapping each other. 
  • feint
    I have decimated muchhh less now, and that big problem area is fine but there are many other black triangles that aren't visible in blender but are coming up in marmoset still :'(

    I used the comment found here to fix some visible areas on the high poly manually, but the areas on the bake aren't visible in blender to me. (shading smooth also seems to bring up many issues with these decimated meshes as mentioned by that commentor)

    I can fix these areas manually in photoshop, but it would be great to understand what's causing it exactly and how to avoid the hassle :p

    Here's a link to the high and low for one part of my scope:
    https://drive.google.com/file/d/1NaIKE1IGwgsJeKWd8TidjTEur7ezuyvy/view?usp=sharing
    https://drive.google.com/file/d/1Tuopzdp43UD5OxiuV5a2b9FGrO65Q21g/view?usp=sharing
  • ZacD
    Online / Send Message
    ZacD ngon master
    Those meshes bake fine in Substance Painter, makes me think it's a baking configuration issue within Marmoset. 
  • Noren
    Offline / Send Message
    Noren polycounter lvl 20
    I just had a quick look at the files themselves and didn't try to bake anything, but the HP has plenty of non-manifolds and degenerate faces, probably courtesy of dynamesh and then carrying over through the decimation process (if not created there). Different software might resolve those differently.
  • Neox
    Offline / Send Message
    Neox godlike master sticky
    Just don't decimate. There is really no need. This isnt gonna be a few hundred million polies, the baker doesnt care.
  • pxgeek
    Offline / Send Message
    pxgeek greentooth
    ZacD said:
    Those meshes bake fine in Substance Painter, makes me think it's a baking configuration issue within Marmoset. 
    since that's the case it could also be gpu/driver related. What gpu are you using? Any issues with cpu baking such as xNormal?
    I took a peek at the HP in max and it looked fine for me.
  • Neox
    Offline / Send Message
    Neox godlike master sticky
    the area looks also fine in toolsbag. but it comes in unsmoothed. on export make sure you export it softened


    and on bake it works for me, just quickly tested in tooldbag 4, ont have 5 installed at the office

    on zero roughness and with an environment shown, i do see some artifacts, which i would assume all come from the hard edged and decimated nature of the HP mesh
Sign In or Register to comment.