well, well, some necro "update" for bmw_m2.zdo:
. (2nd sub mesh)

Remember for the uv start address (see some posts above):
uvaddress= vertexstart + vertexcount x 18
(works for the few models I've checked)

just start with "div 500" (= multiply with 0.002):
.

yeah, my bad, uvs need to be scaled down (div 500 or so), y mirrored, then y scaled up by 3 (round about):
.

Using hex2obj (view link in my sig):
.

So much trouble for nothing. (: Bisai01.perm.bin (starts with 0x6F) works with my tool. But it doesn't care for textures (and was nether intended to do so.) The 0x90 confusion is solved - this is from Mariusz' script: #texture 90 A0 BF CD related files ("header" file (?) and data file): Bi...

well, now I understand the problem - seems an earlier version of the tool extracted obj from Bisai01_ts001.perm.bin for example (starts with 0x90). (But I can't find the concerning exe file - maybe it's on my old PC, somewhere? :cry: ) edit: yeah, no, it was Bisai01 .perm.bin, confused seven years l...

well, 13 vertices, that's what I get. (Iirc there was a threshold required for correct values.)
. edit: gosh, this doesn't seem to be a model file - guess for a header file for the textures (in another file)

Any ideas, trying it on HK.perm.bin and a few others in Zone and Section. That's level files, isn't it? Afair the extractor wasn't intended to be used with them. Only certain vehicles work to add, but 90% of the perm.bins do not. Usually it says "This doesn't seem to be a perm.bin file!" ...

simpler one (vertex and uv data in 24 bytes blocks):
. (First sub mesh only and normals ignored as always by me )

Structure of "blotling"-nif is a little bit different:
.

Getting the mesh directly from that nif (version 20.6.5.0) is not too hard (hex2obj, view link in my sig):
.

Fable 2, GateIron, well, counts are higher, no time for more fiddling:
.

I think I've officially reached the end of my current skillset with this and to be honest it probably would have been quicker and more satisfying making the damn models myself :roll: Rofl! Yep, after 5 years there's some frustration, understandable. I just wanted to give up after 30 minutes of fidd...

well, I have no idea why the face indices don't seem to fit and no time for fiddling, sry:
. Maybe face indices for the head only (0x94d5, 442 vertices) are simply to track?