New Progress on Force Unleashed 2 DLC

[shakotay2]

I checked cpu6.cluster (as advised by zaramot ) and this is the result:

GPU6-CLUSTER.png

H2O file:
0x0 12000
Vb1
32 99
0x5E0C 2166
120000
0x0 255

(uvs not checked so far)

[nightwolf1982]

I think we're a bit off from what I was trying to accomplish here. I'm trying to find a way to extract character models (specifically Jedi Leia) from the larger Endor DLC files. I do appreciate shakotay2's efforts, but I'm not seeing how he's getting the results he's getting.

This is all I've gleaned so far, based on the examples posted and the tutorials I've looked at:

The GPU files appear to store both the face indices and the vertex information for the models (based on shakotay2's last post)

The vertices seem to be coded in hex in the pattern of 3d xx xx xx 3f, though not every set is a vertex

The face index count is roughly the length of the data chunk divided by 2 (again based on shakotay2's last post)

[shakotay2]

I think we're a bit off from what I was trying to accomplish here. I'm trying to find a way to extract character models (specifically Jedi Leia) from the larger Endor DLC files. I do appreciate shakotay2's efforts, but I'm not seeing how he's getting the results he's getting.

Again I don't know what you are talking about.
I checked the files from the 2nd link in your opening post:
https://drive.google.com/file/d/19CpBFO ... tPJTL/view
especially GPU6.CLUSTER.

You can get the head by copying the 6 lines (which I wrote in my previous post) into a text file and rename it to whatever_head.H2O.
Load the GPU6.CLUSTER into hex2obj (File/Open) then File/Open H2O to open the H2O file.
Press the mesh button and the head should be displayed.

(Also keep in mind that it's big endian here, so 00 00 80 3F is 3F 80 00 00 in the files.)

Guess the basic problem is that you don't know how to get the vertices start address. You need some experience for this. There's some trial and error when you try to create point clouds but it's not "magic" especially because the FVFsize is given (32).
(If you don't know what FVFsize is I would suggest to read the tutorial again (tut button in hex2obj) and try out some simple models in the dozens of posts using hex2obj.)

I'm off now; good luck with your further research.

[nightwolf1982]

I finally figured out what I was doing wrong with Hex2Obj, I wasn't using the correct FVF sizes. So, now I've found the face/vertex data for the Endor Leia Costume (Image 1). However, I haven't been able to figure out the UV's yet.

Another problem is that GPU51 seems to be the only file that has mesh information, but the whole file only holds the Endor Leia costume. In the rest of the files that might contain models, each potential model has a section of code that doesn't appear to be vertex data (Image 2) followed by a section that might contain vertex info; but I haven't been able to get any useful point clouds (Image 3).

[nightwolf1982]

second image:

[nightwolf1982]

third image:

[nightwolf1982]

I've uploaded the H2O and GPU files for the Endor Leia costume as well as the files I think have possible model data.

Leia Endor:
https://drive.google.com/open?id=1mL9YH ... j5qJ4RG4BD

GPU Files:
https://drive.google.com/open?id=12mfOp ... aaWBQyFc9N

[shakotay2]

I finally figured out what I was doing wrong with Hex2Obj, I wasn't using the correct FVF sizes.

Very cool that you skipped this hurdle!

So, now I've found the face/vertex data for the Endor Leia Costume (Image 1). However, I haven't been able to figure out the UV's yet.

That's where some experience would come in handy:

GPU51-CLUSTER.png

approach to start of uvs (0x8780C):
start of vertices 0x7F500 + vcount 1676 x FVFsize 20 = 0x877F0

(As I wrote in other threads: there's no general recipe to get the uvs. In most cases they are in the FVF block but here they're not. After having analyzed some dozens of different 3D formats you'll know more.)

[shakotay2]

In GPU16.CLUSTER mesh uses half floats, uvs use shorts ("WordUV"). Search for threads where a workaround for hex2obj is described (since it doesn't provide this combination. You can SaveAs mesh with HF_all selected then switch to WordUV, again SaveAs mesh and copy/overwrite the uvs from that obj to the first one).

GPU16-CLUSTER.png

[nightwolf1982]

Big thanks to shakotay2, I'm now getting meshes from some of the files. Unfortunately, I'm still not finding the one model I'm wanting, Jedi Leia.

I've been taking a close look at the LPMD files, and this is what I've found:

Code: Select all

0x00 = 4c504d44 - LPMD Magic
0x04 = 0000000b - Version?
0x08 = ???? - 0047b000 for all of the Player Outfits
0x0c = ???? - 07dc0000 for the Endor1 and Endor2 files, 00f19999 for all of the Player Outfits
0x10 = ????
0x14 = 00000034
0x18 = 00000001
0x1c = Hex Address 28 bytes from ENUS
0x30 = ????
0x34 = 00000000
0x38 = ffffffc7
0x3c = ????
0x40 = ????
0x44 = Hex Address 67 bytes from STRG / 83 bytes from the string table
0x48 = ????
0x4c = 00000000

0x50 = f0e0d413 - Found throughout two sections of the file, immediately before and after the ENUS section

ENUS - Address varies from file to file

	Unit32 - ENUS
	Unit32 - File Count
	Unit32 - File Number
	Unit32 - ?
	Unit32 - ?
	Unit32 - File Size
	Unit32 - Pad
	Unit32 - Pad
	Unit32 - ?
	Unit32 - ?
	Unit32 - ?

Immediately after the STRG table, there are data blocks for each CLUSTER file, including the full path and name of the file.

[Acewell]

I've been taking a close look at the LPMD files

you should've treated these samples like an archive from the beginning since all the offsets and
names etc are given, break it down to smaller manageable files and it could be easier to work from.
if aluigi had seen these files he could have possibly had an extraction script on the same day.

[nightwolf1982]

The problem I'm having is that while I can partly break down the LPMD file, I'm not experienced enough to figure out how it links out to the CLUSTER files.

For example, I suspect that the hex value f0e0d413 is somehow linked to individual files, but I have no idea how that works. I'm also not certain how the CLUSTER data blocks at the end of the file work.

[nightwolf1982]

Little bit more info

Code: Select all

0x00 = 4c504d44 - LPMD Magic
0x04 = 0000000b - Version?
0x08 = ???? - 0047b000 for all of the Player Outfits
0x0c = ???? - 07dc0000 for the Endor1 and Endor2 files, 00f19999 for all of the Player Outfits
0x10 = ????
0x14 = 00000034
0x18 = 00000001
0x1c = Hex Address 28 bytes from ENUS
0x30 = ????
0x34 = 00000000
0x38 = ffffffc7
0x3c = ????
0x40 = ????
0x44 = Hex Address 68 bytes from STRG / 84 bytes from the string table
0x48 = ????
0x4c = 00000000

0x50 = f0e0d413 - Found throughout two sections of the file, immediately before and after the ENUS section
       
       f934f180 - Found Before and after ENUS; lines up with f0e0d413; Always appears 28 bytes before "GPU" and immediately after "GPU"

      1d9e31ba - Found before and after ENUS; lines up with f0e0d413; Always appears after "LCPU", "SCPU", and "SGPU"

ENUS - Address varies from file to file

	Unit32 - ENUS
	Unit32 - File Count
	Unit32 - File Number
	Unit32 - ? - Can be 0x00 or 0x01
	Unit32 - 0x02
	Unit32 - File Size
	Unit32 - Pad
	Unit32 - Pad
	Unit32 - ?
	Unit32 - ?
	Unit32 - ?

Immediately after the STRG table, there are data blocks for each CLUSTER file, including the full path and name of the file.

[nightwolf1982]

Success!! I've found Jedi Leia! I've managed to extract most of her mesh, but I'm running into an issue with her skirts https://drive.google.com/open?id=1kqVOM ... 53TlxQ02cG

I can get what should be the mesh, but it's full of holes, and I haven't been able to pin down the UVs.

I've uploaded the CLUSTER and h2o files I've got here. https://drive.google.com/open?id=1kbZmj ... D9nkxY7PCS

[shakotay2]

Success!! I've found Jedi Leia! I've managed to extract most of her mesh,

Great!
I always wish that people make progress - seems you're one of them.

Switch to (triangle) Strip(s) for the shirt, uvs are a little bit tricky here (I started with 0x2A80 and increased by 4 until they matched, so just brute force:D)

(6 lines for H2O file):

0x1780 450
Vb1
20 99
0x1B80 188
121000
0x2A8C 56

(One face seems to be missing, though. Dunno why.)