Tekken Hybrid

[gjinka]

So I was right. Thats why Daemon failed.

@howfie: http://www.2shared.com/file/QReIYGqm/nina1u.html

[MrAdults]

So I was right. Thats why Daemon failed.

Quite the contrary, actually. Daemon Tools or any other virtual disc image software should work fine and for most PSX games and will allow you to copy data files straight out of the image. If it's failing then you probably have a screwed up disc image, although it's also possible you've stumbled upon an unusual game without a standard filesystem.

[finale00]

So I got a copy of tekken 3, mounted it, saw some files, but can only look at two of them (SCES_012.37, TEKKEN3.BNS)
The rest are the same situation.

[gjinka]

Quite the contrary, actually ... If it's failing then you probably have a screwed up disc image. filesystem.

It's probably not a screwed up disk because I tried two and finale00 tried himself,

although it's also possible you've stumbled upon an unusual game without a standard filesystem.

And that's what I assumed, doesn't that make my assumption right?

Anyway, it seems somewhat similar to the standard ISO because few small file can be read.
You can have a look at a 4mb slice of the file I posted in the first post if you want. I'm not sure if it's worth it, though, Tekken Hybrid contains Tekken Tag Tournment HD, which has the same models from Tekken 3 (same design), only highpoly.

[howfie]

Almost there... I figured out how the model is divided into surfaces... however... the tristrip format is really strange. For example, nina has the following surface indices:

Code: Select all

size of vertex buffer in bytes = 160000
number of vertices = 5000

size of index buffer in bytes = 27132
number of indices = 13566

sum of surface indices = 13566
indices = 13566
leftover = 0

surface indices = 878
surface indices = 1266
surface indices = 878
surface indices = 577
surface indices = 88
surface indices = 95
surface indices = 31
surface indices = 126
surface indices = 41
surface indices = 577
surface indices = 96
surface indices = 99
surface indices = 115
surface indices = 178
surface indices = 116
surface indices = 743
surface indices = 12
surface indices = 21
surface indices = 154
surface indices = 34
surface indices = 18
surface indices = 14
surface indices = 102
surface indices = 72
surface indices = 26
surface indices = 1251
surface indices = 98
surface indices = 743
surface indices = 110
surface indices = 113
surface indices = 271
surface indices = 162
surface indices = 173
surface indices = 30
surface indices = 186
surface indices = 60
surface indices = 212
surface indices = 1064
surface indices = 34
surface indices = 102
surface indices = 26
surface indices = 33
surface indices = 5
surface indices = 41
surface indices = 36
surface indices = 12
surface indices = 12
surface indices = 28
surface indices = 28
surface indices = 28
surface indices = 28
surface indices = 18
surface indices = 23
surface indices = 38
surface indices = 38
surface indices = 34
surface indices = 30
surface indices = 22
surface indices = 22
surface indices = 317
surface indices = 8
surface indices = 164
surface indices = 20
surface indices = 94
surface indices = 15
surface indices = 36
surface indices = 17
surface indices = 36
surface indices = 16
surface indices = 16
surface indices = 30
surface indices = 4
surface indices = 27
surface indices = 8
surface indices = 40
surface indices = 11
surface indices = 22
surface indices = 9
surface indices = 20
surface indices = 25
surface indices = 28
surface indices = 8
surface indices = 16
surface indices = 14
surface indices = 30
surface indices = 6
surface indices = 40
surface indices = 33
surface indices = 76
surface indices = 8
surface indices = 58
surface indices = 302
surface indices = 186
surface indices = 18
surface indices = 122
surface indices = 18
surface indices = 200

Now this is where things get strange... the first surface, which are her breasts and upper back triangles, says there are 878 indices used, which means 876 triangles total. When rendered as a typical tristrip, all the triangles are there and are perfect, however, it appears that after a degenerate triangle occurs in the tristrip (which is quite common), the winding tends to, but not always, reverses itself, resulting in the following model (degenerate triangles have been removed).



Of course, if I render this as double sided, you see this:



It's really weird... there also appears to be duplicate surfaces... if I take one of them and flip it it seems to generate the right model. Ha ha ha, what a weird, wasteful way of doing this.

[howfie]

Got textures now too... simply just uncompressed headerless DDS. still can't figure out the tri-strip problem however...

[gjinka]

Can you remind me what degenerate triangles are? You say they appear normal when filpped, which sounds like a normals issue.

Good job. Is there anything left in the file. Please post where stuff are (like a wiki) when you get the time.

[MrAdults]

And that's what I assumed, doesn't that make my assumption right?

Nope. That has nothing to do with whether Daemon Tools can successfully mount it. If you can't mount it then you've found quite a rarity amongst PSX games or you have a bad rip. If you can mount it but don't see files representing all of the data, then you simply have a case where the game is accessing additional data from extra-filesystem sectors.

You can have a look at a 4mb slice of the file I posted in the first post if you want. I'm not sure if it's worth it, though, Tekken Hybrid contains Tekken Tag Tournment HD, which has the same models from Tekken 3 (same design), only highpoly.

From what I read, all you posted was the first 4MB of the actual disc image, which isn't particularly useful. I also don't have the disc myself, I only have Tekken Tag Tournament for PS2, which exposes all of the data directly through the standard disc filesystem.

[howfie]

Can you remind me what degenerate triangles are? You say they appear normal when filpped, which sounds like a normals issue.
Good job. Is there anything left in the file. Please post where stuff are (like a wiki) when you get the time.

A degenerate triangle is a triangle that has lost one or more dimensions. A non-degenerate triangle can only be drawn on a plane and its area is never zero. So triangle indices 0-1-2 marks a full triangle. 2-2-3 marks a degenerate one because now you don't have a triangle, you have a line since only two points are used to define the triangle. 3-3-3 is also degenerate since it's point. They commonly occur in tri-strips so they can change direction and skip over spaces. Degenerate triangles are also important in the optimization algorithm of Nelder and Mead.

Yes, I haven't processed the skeleton yet. That is easy too... it's a header section followed by 0x40 byte entries where the first 3 floats per entry are the joint locations and the rest of the data is some kind of matrix. It is the only thing left in the file. Animations are stored somewhere else. These parts comes last. I will update post when I fix the tri-strip problem.

[finale00]

Oh, that's why I see a bunch of repeating indices all over the place.
Is there a common algorithm for processing those kinds of index buffers?

More over, you should add that to the wiki!

[chrrox]

here is an example loop in max its pretty easy to follow.

StartDirection = -1
f1 = (ReadBEword f)
f2 = (ReadBEword f)
FaceDirection = StartDirection
while (ftell f) < (END) Do (
f3 = (ReadBEword f)
if (f3==0xFFFF) then (
f1 = (ReadBEword f)
f2 = (ReadBEword f)
FaceDirection = StartDirection
) else (
f3 += 1
FaceDirection *= -1
if (f1!=f2)AND(f2!=f3)AND(f3!=f1) then (
if FaceDirection > 0 then append Face_array [(f1),(f2),(f3)]
else append Face_array [(f1),(f3),(f2)]
)
f1 = f2
f2 = f3
)
)

[gjinka]

Nope. That has nothing to do with whether Daemon Tools can successfully mount it. If you can't mount it then you've found quite a rarity amongst PSX games or you have a bad rip.

What? You are saying the same thing as last time.
I can't add anything more to what I've said:

Quite the contrary, actually ... If it's failing then you probably have a screwed up disc image. filesystem.

It's probably not a screwed up disk because I tried two and finale00 tried himself,

although it's also possible you've stumbled upon an unusual game without a standard filesystem.

And that's what I assumed, doesn't that make my assumption right?

If you can mount it but don't see files representing all of the data, then you simply have a case where the game is accessing additional data from extra-filesystem sectors.

"Extra filesystem sector"?

[howfie]

Ok just panning out some thoughts... this is what's going on... all of a sudden the tristrip draws flipped triangles. It's not clear what the reset condition is because sometimes it does this after degenerates, but sometimes not. In the example above, the strip flips after 4 degenerates. It looks like 3 in the picture but LightWave says there's 4 so . Here's the data:

Code: Select all

00 00 01 00 02 00
03 00 04 00 05 00
06 00 07 00 08 00
09 00 0A 00 0B 00
0C 00 0D 00 0E 00
0F 00 10 00 11 00
12 00 13 00 14 00
15 00 16 00 17 00
18 00 19 00 1A 00
1B 00 1C 00 1D 00
1E 00 1F 00 20 00
21 00 21 00 22 00
22 00 23 00 02 00
24 00

degenerate: 20 00 21 00 21 00
degenerate: 21 00 21 00 22 00
degenerate: 21 00 22 00 22 00
degenerate: 22 00 22 00 23 00
strip flip: 22 00 23 00 02 00

Now let's move on to the next strip flip occurrence and see if something is in common. Well what do you know... it doesn't flip after a degenerate this time!



So what's the data look like? As you can see... four consecutive degenerates again, but this time, the strip doesn't flip.

Code: Select all

3D 00 3E 00 3F 00
40 00 40 00 41 00
41 00 42 00 43 00
44 00 45 00 46 00

degenerate: 3F 00 40 00 40 00
degenerate: 40 00 40 00 41 00
degenerate: 40 00 41 00 41 00
degenerate: 41 00 41 00 42 00
stip noflip: 41 00 41 00 42 00

So there's a dead end there. But as I mentioned before, the chest area geometry appears twice in surface #0 and in surface #2. Here is the surface data for those two things at the beginning of the file. Notice the duplicate 27228 values. That appears to be an identifier for the chest. Sometimes that identifier value is 0 which means that the surface doesn't appear more than once. This is the case for her arms.

Code: Select all

surface 0
0
0
68 (always 68)
27228 (chest identifier)
3
0
0
0
0
2
0 (texture identifier)
65537
0 (offset)
878 (number of indices in surface)

Code: Select all

surface 2
1
1
68 (always 68)
27228 (chest identifier)
0
0
0
0
0
2
2 (texture identifier)
65536
2144 (offset)
878 (number of indices in surface)

The only real difference between surface 0 and surface 2 is that they use different textures. The number 3 after the chest identifier probably indicates surface 0 is some type of effect surface since the texture appears to be some kind of lightmap or something? Not sure.

Surface 0 uses:


Surface 2 uses:


So we're back to square one... each surface should draw the geometry correctly without the problem with the strips. It's definitely not a front/back surface type of thing. So guess what? It is time to give up lol. I'm going to conclude that this game uses double-sided polygons and that the triangle strips are fine. However, modelers like lightwave and maya are going to only use the vertex normals from the OBJ export for lighting and not for polygon direction. So in other words, when you load the model into a modeler you are going to see crap if using OBJ.

[howfie]

Fixed! Had to write a function that flips triangles if computed normals do not match up with the average of the vertex normals. Almost ready for release.

[finale00]

Nice.

What complicated face parsing