Hi Resric,

It sounds like you may have the wrong wrap direction for the input
triangles.  Wrap direction is how NMGen detects the up-face.  The following
page includes a helpful visualization:

http://www.critterai.org/projects/cainav/doc/html/8a99e478-a235-49ef-...

If that isn't the problem, could you send me your configuration?  I can
check it for common issues.

Also, if you have a link to the page with the documentation errors I'll
take a look and get it fixed.

-Steve

I hadn't seen that page so that is definitely my problem (at least one of
them) since my polys are CCW. I had suspected as much and tried simply
reversing the order of the indexes on all my polys, but that left me no
detail mesh. Is that the correct approach to change my winding?

http://www.critterai.org/nmgen_config

That link is probably correct as part of the Java study, but as a C#
reference the reader has a lot of matching up to do. Some of the parameters
don't seem to exist for C# and some have different types which seem to
contradict the suggested values I've seen. I've been trying to glean what I
can from the Unity sample but having never worked with the engine it is
slow going.

It all depends on your source format. Sometimes it's just a matter of
swapping indices.  Other times you'll need to flip an axis (usually the
x-axis) by multiplying its vertex values by -1. I often revert to good old
graph paper to help me figure out what to do when dealing with an
unfamiliar format.

That linked configuration document only applies to the Java study, which I
don't keep in synch with CAINav.  CAINav is moving forward while the Java
Study is pretty much frozen. The CAINav
documentation<http://www.critterai.org/projects/cainav/doc/Index.html>contains
an
*Introduction to NMGen -> Recommendations* page that provides some extra
information related to choosing configuration settings.

-Steve