straaaaaaange normal maps + tangents
polycounter
Hey ho,
here's a short summary of what I'm into but what I'm asking is more a low-level-question
I'm exporting a model from 3ds max with my own exporter to gta4-community-open-format this get's through openIV into gta4
Now my Problem:
All parts of my model where the uvs are (more or less) rotated 90° to the left are looking all right.
BUT - all parts that are oriented normally look wrong. And all normal map channel swapping and flipping can't fix that.
ONLY if I rotate the whole uvs 90° to the left (+ the normal map of course) I come to a good result (some channel flipping might be necessary)
Of course now the parts that looked good before are now like wtf!?
My Question:
What kind of tangent calculation stupidness could cause such a behavior?
I use this method from 3ds max btw
http://area.autodesk.com/blogs/chris/how_the_3ds_max_scanline_renderer_computes_tangent_and_binormal_vectors_for_normal_mapping
which gives me really good results on the -90° parts
here's a short summary of what I'm into but what I'm asking is more a low-level-question
I'm exporting a model from 3ds max with my own exporter to gta4-community-open-format this get's through openIV into gta4
Now my Problem:
All parts of my model where the uvs are (more or less) rotated 90° to the left are looking all right.
BUT - all parts that are oriented normally look wrong. And all normal map channel swapping and flipping can't fix that.
ONLY if I rotate the whole uvs 90° to the left (+ the normal map of course) I come to a good result (some channel flipping might be necessary)
Of course now the parts that looked good before are now like wtf!?
My Question:
What kind of tangent calculation stupidness could cause such a behavior?
I use this method from 3ds max btw
http://area.autodesk.com/blogs/chris/how_the_3ds_max_scanline_renderer_computes_tangent_and_binormal_vectors_for_normal_mapping
which gives me really good results on the -90° parts