Home Adobe Substance

Bakeing hell

Hellgnome
polycounter lvl 8
Offline / Send Message
Hellgnome polycounter lvl 8
Hi all, 

So i am getting these weird maps when baking, and I´d like to know, if anyone knows why this usually happens?

It is an ambient occ map and a defuse map btw.


Replies

  • Fabi_G
    Offline / Send Message
    Fabi_G high dynamic range
    Hi! Some more context would help to understand, like what does the mesh looks like? Ideally share files.
  • Hellgnome
    Offline / Send Message
    Hellgnome polycounter lvl 8
    Fabi_G said:
    Hi! Some more context would help to understand, like what does the mesh looks like? Ideally share files.

    How do I share the file? There is no option to add the file here on the forum? :)

  • Fabi_G
    Offline / Send Message
    Fabi_G high dynamic range
    Hm, it should be this:

    With small files, you can zip them and attach directly to a post. With larger ones, I would host them somewhere (drive, Dropbox,  ...) and share the link.
  • Hellgnome
    Offline / Send Message
    Hellgnome polycounter lvl 8

    Oh! Could not share them as FBX, didn´t even think to pack them :D 
    Thanks :) 
  • Fabi_G
    Offline / Send Message
    Fabi_G high dynamic range
    Hi! Doing a bake with the attached files in Painter, didn't reveal the issues you show in your original post. The file contains just a single door element. Maybe the issue you presented is caused by the way UVs are stacked and due to the bake settings? Did you bake with your test file as well and reproduce the issue?

    That aside, here are some notes/opinions from looking at the model:

    - Separating repeating modules (hinges) and instancing them results in less unwrapping and more available pixels. Using bake group / match by name functionality prevents projection errors. Offset UVs of instances by 1 so they're out of the 0-1 space.

    - Sharp 90 degree angles of the high poly will not show up in the normal map.

    - With an average mesh normals projection, depending on the geometry, floating elements (inside of door) will be distorted due to the skewing that occurs. To get a clean result here, you could:
    • modify your lowpoly geometry so that floating parts don't distort
    • composite a mesh normals bake with an average normals bake (edges). Toolbag would allow to control the skewing by a mask.
    • simply bake down the cabinet elements on an atlas, then map the game meshes UVs to them

    Or match lowpoly and highpoly more closely and bake with average mesh normals. Probably the most straightforward high to low-poly way.
  • Hellgnome
    Offline / Send Message
    Hellgnome polycounter lvl 8
    Fabi_G said:
    Hi! Doing a bake with the attached files in Painter, didn't reveal the issues you show in your original post. The file contains just a single door element. Maybe the issue you presented is caused by the way UVs are stacked and due to the bake settings? Did you bake with your test file as well and reproduce the issue?

    That aside, here are some notes/opinions from looking at the model:

    - Separating repeating modules (hinges) and instancing them results in less unwrapping and more available pixels. Using bake group / match by name functionality prevents projection errors. Offset UVs of instances by 1 so they're out of the 0-1 space.

    - Sharp 90 degree angles of the high poly will not show up in the normal map.

    - With an average mesh normals projection, depending on the geometry, floating elements (inside of door) will be distorted due to the skewing that occurs. To get a clean result here, you could:
    • modify your lowpoly geometry so that floating parts don't distort
    • composite a mesh normals bake with an average normals bake (edges). Toolbag would allow to control the skewing by a mask.
    • simply bake down the cabinet elements on an atlas, then map the game meshes UVs to them

    Or match lowpoly and highpoly more closely and bake with average mesh normals. Probably the most straightforward high to low-poly way.
    Thank you so much for all the tips, that is deffinetly something I will use in the future.

    In regards to reproducing the problem, I don´t know, we ended up making a completely new model, and the problem never returned.
Sign In or Register to comment.