It might be useful if you can upload a sample of 1 or 2 files.

Firstly, download QuickBMS: http://aluigi.altervista.org/quickbms.htm - save my bms script somewhere you'll remember. I usually put them in the same folder as the data file to be extracted. If you run QuickBMS from the exe, it'll ask you to select the script, then the file (bgm.dat), then an output ...

If you use my attached QuickBMS script to extract the FSB file, you can then drop that extracted FSB file straight into Foobar and vgmstream can play/convert the files as required.

The index to the files is stored inside the .xmb files (which weren't in the uploaded data), and then it all becomes a bit clearer. So in objects-e01.xmb, for example, at offset 0xbfb you've got 0x290 for the faces and 0x26 vertices at 0xb8acc. Seems like each entry is 0x39 bytes. Parts of these xmb...

It's certainly an odd format. While it's easy to get the data for the vertices, UVs and faces, there doesn't seem to be anything obvious referring to the actual mesh parts. In the case of the MountCar file, there are several tables after the face data, each with 0x44 entries - I can see tables for w...

I hadn't extracted that movies.mgf previously, but it does have some .xsb files that match those .hdxwb files - so you can get the filenames for the sfxstream and voice - if you rename those other .hdxwb files to just .xwb you can drop them straight into Foobar instead. I'd already written my script...

The .xwb files are fine. You can extract mgf archives with my attached script, and then you can drop the xwb files into Foobar to get the sounds.

It looks to me that there's a couple of issues with the .wav file: Firstly, the codec is listed as 0x69 - it should be 0x165 for XMA, which is why you get "XMA code not found" error. Secondly, the header is shorter than it should be - for XMA the header should be 0x3c bytes, where the audio data sta...

Yes, use this script and the attached .txth file for the sfxstream file.

Here's an initial basic Noesis script that should read the vertices and faces correctly from the text info supplied in the file.

You'll need to rename the files as *.bingeom for the script to work.

Some more info: if you look at the end of the file, there is some text info that tells you how the data is stored. So in the attachment in the first post, you have this: {"type":"PolyMesh","faces":{"o":0,"l":1092,"b":1},"faceCount":{"o":1092,"l":4},"positions":{"o":1096,"l":2184},"colorMaps":{},"uvM...

This is my script so far, if you want to test it out. Not sure if it's perfect, but hopefully it should work with most models.

Here's my crappy initial Noesis script which should get the basic model data, with the faces. The file probably needs extensive parsing of the PS2's VIF data to get the full thing, for UVs and normals, etc. It's not perfect, but it's a start... It works similar to that Blender one, just as a test of...

Can you upload 1 or 2 examples of the model files?

I've not looked in much detail, but I can see this object in one of the .mesh files. The data in those seems to start at 0x15c and each entry is 0x3c bytes (5038 vertices in this case). I can see vertices and UVs, not sure about most of the rest. The .skin file contains all the skeleton info, and .m...