I know NifUtilities is designed to work with nif files created during export from a 3D modeling program, specifically 3dsmax and possibly Blender 2.49b. I have however come up with a workflow for non-rigged models(static, clutter and weapon) in Blender 2.65(just can't handle 2.49b interface as it's atrocious).

I'm trying to figure out exactly what Chunk merge will accept as a collision model.

Let me first explain my work flow.

1} I'm using Blender 2.65 to model custom shapes and modify existing shapes.

2}To get a new shape into a Skyrim .nif
a}I am exporting the shape from Blender 2.65 as a .3ds
b}and then importing over an existing shape in a Skyrim .nif.

3} To modify an existing shape in a Skyrim.nif
a)I first delete the collision branch and then export as a .dae
b}Import dae file to blender
c}alter shape then proceed with 2} listed above

My attempt to integrate Nifutilities into my work flow.

1}Opened Nif convert.exe
2} selected valid skyrim mesh with no collision
3} selected temp.nif as template file
4} resulting nif contains necessary collision branch BUT it duplicated the NiTrishape
5}Opened Chunkmerge.exe
6}selected nif.convert output nif(from step 4) as source file
7}selected original valid skyrim mesh(from step 2) containing shape and no collision as collision file.
8} resulting mesh contains a valid collision BUT It is somewhat simplified and more or less incomplete compared to the original geometry of the shape.

Any information you can give me to help refine this process would be greatly appreciated. If you would like to see my file(s) i will upload them just specify exactly what you want to see or all if that's necessary.

Anything you can tell me to further prep the files before processing would be great. For instance: Should the collision file contain nothing more than a shape with no BSFade node or BSLightingProperty?