Home Technical Talk

Normal Map Seams between UV Shells

polycounter lvl 5
Offline / Send Message
Panzerdraco polycounter lvl 5
Having an issue baking a tangent normal map.

The below image is a normal map baked from 3DS Max and rendered in Marmoset.
1. The lowpoly's smoothing groups are split by UV shell.
2. I am using a projection cage.
3. Gamma correction is disabled.
4. Marmoset's tangent space is set to the appropriate program match.

50589c7a72.jpg

Additionally, here's another bake done using xNormal, using the same input models and projected cage. You can see the issue is almost identical.

1c1d3ed34f.jpg

Any ideas what could be causing that seam?

Replies

  • Mossbros
    Offline / Send Message
    Mossbros polycounter lvl 9
    a bit of a shot in the dark, but are your uv's rotated at all ?

    like instead of this :- l8zJTDv.png

    You have this :- RJPirnB.png
  • Panzerdraco
    Offline / Send Message
    Panzerdraco polycounter lvl 5
    3b652af356.png

    That would be a negative!
  • Panzerdraco
    Offline / Send Message
    Panzerdraco polycounter lvl 5
    1c31517a82.png

    Additionally, here's a photo of my baking settings.
  • Mossbros
    Offline / Send Message
    Mossbros polycounter lvl 9
    been messing around trying to find out what's up since 2am lol, got started and remembered when I had the same issue in Xnormal.

    Several things it could be, in no particular order.
    1. lowpoly normals/sg haven't exported with the obj (this one I've had issues with previously as maya can sometimes forget things.)
    2.padding issue (unlikely)
    3.mesh scale in xnormal crank it up to 20 on all the meshes
    4. the tangent gods are angry with the lack of your sacrifices.

    I'm sure Lee Devonald see's people post the same problem daily and it's something pretty simple to fix in all honesty, but my brain is chugging at the moment to remember.
  • alifarsangi
    Offline / Send Message
    alifarsangi polycounter lvl 12
    u might try this. reset xform on ur mesh.
  • Panzerdraco
    Offline / Send Message
    Panzerdraco polycounter lvl 5
    http://panzerdraco.net/FTP/NovasurgeTest.zip
    If it helps, I've uploaded my HP, LP and cage for you guys to play with.
  • Panzerdraco
    Offline / Send Message
    Panzerdraco polycounter lvl 5
    Mossbros:
    1. Checked to make sure that this wasn't the case. Not the issue.
    2. Tried baking with more padding, still an issue.
    3. This made no difference.
    4. I'm considering this as a possibility.

    alifarsangi:
    Didn't make a difference.
  • huffer
    Offline / Send Message
    huffer interpolator
    That's how hard edges will look like baked, and zoomed in...the only fix I've found is baking in Maya, which gets them much better but still not 100 percent. Here's what I found.
  • Panzerdraco
    Offline / Send Message
    Panzerdraco polycounter lvl 5
    huffer wrote: »
    That's how hard edges will look like baked, and zoomed in...the only fix I've found is baking in Maya, which gets them much better but still not 100 percent. Here's what I found.

    Wow, this is some great info. I'll try the map dilation in PS first.
  • cupsster
    Offline / Send Message
    cupsster polycounter lvl 11
    Bleed the edges for sure! A lot of artefacts at edges is caused from not bleed pixels around islands. Use "Solidify C" from flaming pearls. Free photoshop plugin package. There are A and B presets but for me best works C preset.
  • Panzerdraco
    Offline / Send Message
    Panzerdraco polycounter lvl 5
    cupsster wrote: »
    Bleed the edges for sure! A lot of artefacts at edges is caused from not bleed pixels around islands. Use "Solidify C" from flaming pearls. Free photoshop plugin package. There are A and B presets but for me best works C preset.

    So I tried both this and the xNormal dilation plugin. Both of them seem to be working fine but the seam remains.
  • ghaztehschmexeh
    You won't notice this when your model is viewed from a normal distance and textured. If it's meant to be seen from this distance then you should probably just use more geometry. Otherwise, you won't notice.

    Here is an example

    eeb56526b1.jpg
    d07abd75aa.jpg
  • Panzerdraco
    Offline / Send Message
    Panzerdraco polycounter lvl 5
    If it's not possible to fix this, then I can accept that answer.
    Thanks for the help, guys.
  • Dan Powell
    Offline / Send Message
    Dan Powell polycounter lvl 5
    Have you tried resetting the normals on both the low and high poly inside 3DS Max?

    Apply a "Normals" modifier once, and check the "Unify" normals button.

    Then do the same again once more, to flip them back out again.

    I don't know why this worked, but I've had very similar results to this before and I did an array of things to try and "quickfix" it, and resetting the normals like above was one of them.

    I also tried resetting the X-Form, but someone else has already suggested that.
  • Panzerdraco
    Offline / Send Message
    Panzerdraco polycounter lvl 5
    Dan Powell wrote: »
    Have you tried resetting the normals on both the low and high poly inside 3DS Max?

    Apply a "Normals" modifier once, and check the "Unify" normals button.

    Then do the same again once more, to flip them back out again.

    I don't know why this worked, but I've had very similar results to this before and I did an array of things to try and "quickfix" it, and resetting the normals like above was one of them.

    I also tried resetting the X-Form, but someone else has already suggested that.

    6b7bee0a7b.png

    I assume this is what you mean? Trying this now.

    EDIT: Made no difference.
  • ActionDawg
    Offline / Send Message
    ActionDawg greentooth
    This may just be due to the limit of what the normals can provide. If you consider that the orientations between normals of two separated faces is going to be discretized to a range of values that can be stored in a 8 bit per channel texture then it makes sense to see small seams where the values can't quite match up. This coupled with the resolution of the pixels themselves have minimal artifacts like this.
  • Prime8
    Offline / Send Message
    Prime8 interpolator
    somedoggy wrote: »
    This may just be due to the limit of what the normals can provide. If you consider that the orientations between normals of two separated faces is going to be discretized to a range of values that can be stored in a 8 bit per channel texture then it makes sense to see small seams where the values can't quite match up. This coupled with the resolution of the pixels themselves have minimal artifacts like this.

    I thought the same.
    You could increase the texture size and see if it reduces the artifacts, for testing.

    Could be caused by split edges on the cage as well.

    Did you test your baking process with a simple model before? Simple cube with round edges works for that.
Sign In or Register to comment.