Difference between revisions of "FF8/FileFormat X"

From Final Fantasy Inside
< FF8
Jump to navigation Jump to search
my_wiki>MaKiPL
(Quad)
m (0 Default: fix size of animation frame)
 
(62 intermediate revisions by 5 users not shown)
Line 1: Line 1:
 
By MaKiPL. Thanks for help in research for: shakotay2 (XeNTaX), Halfer, Yagami Light.   
 
By MaKiPL. Thanks for help in research for: shakotay2 (XeNTaX), Halfer, Yagami Light.   
TODO: Update with CLUT (which byte)
+
Complete list of original battle stages by Kaspar01: [[FF8/BS_list|List of battle stages]]
TODO: Check, if entry data has something to do with TimeCompression (or if it's engine handled)
 
TODO: Do that camera finally!
 
  
 
== Info ==
 
== Info ==
  
.X file is uncompressed 3D stage model with texture embedeed. The file contains basic info (currently unknown, probably for editor?), camera data, movement, translations and geometry data.  
+
.X file is uncompressed 3D stage model with texture embedeed. The file contains unused info (not used by FF8 engine), camera data, movement, translations and geometry data.  
 +
 
 +
Battle stages '''DOES NOT''' contain pointers to next sections. All pointers are '''HARDCODED''' in FF8.EXE. [[FF8/BattleStage/Pointers|Click me for battle stage pointer list]]
  
 
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
Line 13: Line 13:
 
! style="background:rgb(204,204,204)" | Description
 
! style="background:rgb(204,204,204)" | Description
 
|-
 
|-
| Info/Script
+
| PlayStation MIPS assembly
 
| E8 FF BD 27 01 00 02
 
| E8 FF BD 27 01 00 02
| Currently unknown. Probably header + info, that is not needed in game itself. Can be null'ied and nothing will happen ingame.
+
| Used only in PS version; skipped in PC
 
|-
 
|-
 
| Camera data + movement
 
| Camera data + movement
| 02 00 08 00 20 00 (usually 0x5d4/8)
+
| 02 00 08 00 20 00
 
| Camera animations, movement (pre-keyed)
 
| Camera animations, movement (pre-keyed)
 
|-
 
|-
| Geometry
+
| Model section
| 01 00 01 00 (See geometry section info)
+
| 06 00 00 00 (See geometry section info)
 
| Divides to object (group) and it's sub-objects + vert/triangle grouping
 
| Divides to object (group) and it's sub-objects + vert/triangle grouping
 
|-
 
|-
 
| Texture
 
| Texture
| 01 00 00 00 09 (Always)
+
| 10 00 00 00 09 (Always)
| Contains one .TIM texture  
+
| Contains one .TIM texture, 8 BPP
 
|}
 
|}
  
== Info/Script ==
+
== How is this file handled by the engine? ==
 +
 
 +
FF8 loads specific file and reads it from hardcoded (written in FF8 code) position (or reads MIPS PlayStation assembly that contains battle stage load code). This hardcoded position points to camera data (in PC). Camera data has size uint16, which is calculated and relative jump is made. Just after camera, a typical section based file handling is made.
  
Currently unknown, takes up to ~1400 bytes from start. Unused in-game. Can be null'ied and nothing happen.
+
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint32
 +
| Number of sections
 +
|-
 +
| 4
 +
| uint32
 +
| Objects group #1
 +
|-
 +
| 8
 +
| uint32
 +
| Objects group #2
 +
|-
 +
| 12
 +
| uint32
 +
| Objects group #3
 +
|-
 +
| 16
 +
| uint32
 +
| Objects group #4
 +
|-
 +
| 20
 +
| uint32
 +
| Texture [unused in code]
 +
|-
 +
| 24
 +
| uint32
 +
| Texture
 +
|-
 +
| 28
 +
| uint32
 +
| Relative to EOF
 +
|}
 +
 
 +
====Objects Group====
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint32
 +
| Number of sections
 +
|-
 +
| 4
 +
| uint32
 +
| Pointer to Settings #1
 +
|-
 +
| 8
 +
| uint32
 +
| Pointer to ObjectsList
 +
|-
 +
| 12
 +
| uint32
 +
| Pointer to Settings #2
 +
|-
 +
| 16
 +
| uint32
 +
| Relative to end_of_group
 +
|}
  
This header accesses opcodes related to Direct3D "shadow" stuff. Is highly connected to AF3DN.p
+
=====Objects List=====
OpCode that accesses this portion of data is "unsigned divide by 2".
+
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint32
 +
| Number of objects
 +
|-
 +
| Num_of_obj*4
 +
| uint32
 +
| Relative pointer to segment (one .obj file) [01 00 01 00]
 +
|}
 +
 
 +
=====Settings 1=====
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint16
 +
| Unknown, fixed 2?
 +
|-
 +
| 2
 +
| sint16 (probably, or uint16)
 +
| Scale (01 is the smallest)
 +
|-
 +
| 4
 +
| 4 bytes
 +
| PADDING
 +
|-
 +
| 8
 +
| 4 bytes
 +
| Nothing?
 +
|-
 +
| 12
 +
| 4 bytes
 +
| Nothing?
 +
|-
 +
| 16
 +
| 2 bytes
 +
| if not FF FF, then model vanishes
 +
|-
 +
| 18
 +
| sint16
 +
| X or Z Axis rot origin (Translation)
 +
|-
 +
| 20
 +
| 44 bytes
 +
| Left null for storing some engine data AFTER loaded in-game
 +
|-
 +
| 64
 +
| 2 bytes
 +
| if not FF FF or 00 00, then model vanishes
 +
|-
 +
| 66
 +
| 2 bytes
 +
| Nothing?
 +
|-
 +
| 68
 +
| ...up to end
 +
| Left null for storing some engine data AFTER loaded in-game
 +
|}
 +
 
 +
This might look difficult, but it's like multi-pointer file that starts at different location given by executable. See list at the beginning of this wiki to see which stages camera starts at which position (where FF8.exe really starts to read file as BattleStage) [FF8 stores whole file, even with useless data before camera]
 +
If this still makes trouble, see this video: TODO
  
 
== Camera data ==
 
== Camera data ==
 +
 +
Starts at ~0x5d4 (see How the engine handles this file?).
 +
[.text:00509820]
  
 
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
Line 45: Line 179:
 
|-
 
|-
 
| 0
 
| 0
| 2 bytes
+
| uInt16
| Number of cameras
+
| Pointers count. Fixed (02 00) '''[Unused in code]'''
 +
|-
 +
| 2
 +
| uInt16
 +
| Relative jump to CameraSetting. Should be (08 00)
 +
|-
 +
| 4
 +
| uInt16
 +
| Relative jump to CameraAnimationCollection. Should be (20 00)
 +
|-
 +
| 6
 +
| uInt16
 +
| Camera data size (starting from 0) '''[unused in code]'''
 +
|-
 +
| 8
 +
| 24 bytes
 +
| Camera Settings
 +
|-
 +
| 32
 +
| ?? bytes
 +
| Camera Animation Collection
 +
|}
 +
 
 +
====Camera Setting====
 +
 
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| char (based on disassembly)
 +
| CameraAnimMode?
 +
|-
 +
| 1
 +
| byte?
 +
| StopEnemyBeforeAnim?
 +
|-
 +
| 2
 +
| Bitfield? (1B)
 +
| UHM?
 +
|-
 +
| 3
 +
| ???
 +
| DefaultEndofCameraPosition_Zoom?
 +
|-
 +
| 4
 +
| ???
 +
| ???
 +
|}
 +
 
 +
====Camera Animations Collection====
 +
 
 +
 
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint16
 +
| NumOfSets
 +
|-
 +
| 2 *(Set index)
 +
| uint16
 +
| Relative pointer to camera Animation Set
 +
|-
 +
| 2 *(Set index)+ 2
 +
| uint16
 +
| Camera EOF
 +
|-
 +
| CameraAnimationSetPointer
 +
| CameraAnimationSet
 +
| CameraAnimationSet
 +
|}
 +
 
 +
=====CameraAnimationSet=====
 +
Seems to be 8 pointers in each set.
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
|0
 +
|ushort
 +
|AnimPointer*2
 +
|-
 +
|AnimPointer
 +
|Varies 144-146-148 bytes usually
 +
|CameraAnimation
 +
|}
 +
 
 +
== Camera Animation (WIP) ==
 +
All this is based on best effort knowledge from reversing by Maki. I'm just trying to understand it.
 +
=== Control Word ===
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint16_t (bit varies)
 +
| Main controller. If 0xFFFFU END
 +
|}
 +
==== FOV ====
 +
Control Word & 0b0000'0000'1100'0000U
 +
===== 1 Default =====
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| None
 +
| None
 +
| Start = 0x200
 +
|-
 +
| None
 +
| None
 +
| End = 0x200
 +
|}
 +
===== 2 Same =====
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint16_t
 +
| Start
 +
|-
 +
| 2
 +
| uint16_t
 +
| Padding
 +
|-
 +
| None
 +
| None
 +
| End = Start
 +
|}
 +
===== 2 Different =====
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint16_t
 +
| Start
 +
|-
 +
| 2
 +
| uint16_t
 +
| Padding
 +
|-
 +
| 4
 +
| uint16_t
 +
| End
 +
|}
 +
 
 +
==== ROLL ====
 +
Control Word & 0b0000'0011'0000'0000U
 +
===== 0 Unknown =====
 +
TODO
 +
===== 1 Default =====
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| None
 +
| None
 +
| Start = 0x000
 +
|-
 +
| None
 +
| None
 +
| End = 0x000
 +
|}
 +
===== 2 Same =====
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint16_t
 +
| Start
 +
|-
 +
| None
 +
| None
 +
| End = Start
 +
|}
 +
===== 2 Different =====
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint16_t
 +
| Start
 +
|-
 +
| 2
 +
| uint16_t
 +
| End
 +
|}
 +
 
 +
==== LAYOUT ====
 +
Control Word & 0b0000'0000'0000'0001U
 +
===== 0 Default =====
 +
You'll loop through the data till the first value is less than 0. pushing back to a variable length container.
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| int16_t
 +
| if < 0 break; could be related to time of frame.
 +
|-
 +
| 2
 +
| 16 bytes
 +
| Animation Frame
 +
|}
 +
 
 +
===== 1 Other =====
 +
TODO
 +
===Time===
 +
Time is calculated from number of frames. You basically set starting position World+lookat and ending position, then mark number of frames to interpolate between them. Every frame is one draw call and it costs 16. Starting time needs to be equal or higher for next animation frame to be read; If next frame==0xFFFF then it's all done.
 +
=== Animation Frame ===
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | Offset
 +
! style="background:rgb(204,204,204)" | Length
 +
! style="background:rgb(204,204,204)" | Description
 +
|-
 +
| 0
 +
| uint16_t
 +
| is frame durations shot
 
|-
 
|-
 
| 2
 
| 2
| 2 bytes
+
| Vertice<uint16_t> (x,y,z)
| Number of camera animations?
+
| World
 +
|-
 +
| 8
 +
| uint16_t
 +
| is frame ending shot
 
|-
 
|-
| ?? / Camera data
+
| 10
| Varies
+
| Vertice<uint16_t> (x,y,z)
| Currently unknown
+
| Look At
 
|}
 
|}
  
Camera data contains:
+
=====Dependency=====
# Prekeyed camera movement
+
 
# Camera zoom
+
  CAMERA:
# Camera FOV (horizontal and vertical)
+
  -Camera Settings
# Orthogenical projection
+
  -Camera Animation Collection (even three collections in a0stg006.x and up to 7 in a0stg101.x!)
# Rotation
+
      |
# Special effects like earthquake
+
      |
# More, unknown?
+
      -Camera animation Set (Always 8 camera animations, may be empty (0xFFFF); look for example to a0stg127.x collectionId==1; there are 8 animations, but 7 of them are 0xFFFF (pointers increase by 2))
This file does not contain camera animations for victory, summons and etc.  
+
          |
 +
          |
 +
          - Camera animation
 +
 
 +
 
 +
=====Example=====
 +
 
 +
a0stg006.x:<br>
 +
0x5d8 -> 02 00 08 00 20 00
 +
# Get EOF-> *(0x5d8 + 8) -> 32
 +
# Jump to EOF -> 0x5d8 + 32 = 0x5F8 (This is now Camera Animation Collection)
 +
# Get pointer to correct anim collection. In this case we will use AnimCollectionID == 0, so:  *(0x5F8 + 0*2 + 2) -> 0x0c
 +
# Jump to Anim collection data: 0x5F8 + 0x0c = 0x604 (This is now Camera Animation Set)
 +
# Jump to Camera animation by cameraAnimSetID, let's take for example cameraAnimSetID == 7, so: *(0x604 + 7*2) -> 0x25E. Now carefully, jump by multiplying it by 2!
 +
## 0x604 + (0x25E * 2) = 0xAC0
 +
 
 +
Therefore: a0stg006.x Camera animation 7 in camera collection 0 is at 0xAC0
  
 
== Geometry ==
 
== Geometry ==
Line 90: Line 478:
 
|-
 
|-
 
| 6 + (Number of vertices *6)
 
| 6 + (Number of vertices *6)
| (AbsolutePosition MOD 4) + 2
+
| (AbsolutePosition MOD 4) + 4
 
| Padding
 
| Padding
 
|-
 
|-
Line 274: Line 662:
 
0x0E
 
0x0E
 
0x05
 
0x05
Other way is to treat first one as 6 BIT's, and second one as 2 BIT.
 
  
 
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
Line 288: Line 675:
 
| B2 --> 02
 
| B2 --> 02
 
| Texture page number
 
| Texture page number
 +
|}
 +
 +
==== CLUT ID ====
 +
 +
The most important bit's are the first two on one byte, and last two on last byte. Example:
 +
00000000 00111100 (00 3C)
 +
should be read like this:
 +
00111100 00000000 (3C 00)
 +
And the CLUT ID is revealed by watching this four bits:
 +
001111[00 00]000000
 +
 +
Same applies to reading CLUT colors (16bits) where they are: 1b-T, 5b-B, 5b-G, 5b-R (Or RGBT if reordered as shown above)
 +
 +
Complete list (without replacing bit order - as is in HEX editor/memory):
 +
 +
{| border="1" cellspacing="1" cellpadding="3" align="center" style="border: 1px solid black; border-collapse: collapse;"
 +
! style="background:rgb(204,204,204)" | CLUT ID
 +
! style="background:rgb(204,204,204)" | BIT
 +
! style="background:rgb(204,204,204)" | HEX
 +
|-
 +
| #1
 +
| 00000000 00111100
 +
| 00 3C
 +
|-
 +
| #2
 +
| 01000000 00111100
 +
| 40 3C
 +
|-
 +
| #3
 +
| 10000000 00111100
 +
| 80 3C
 +
|-
 +
| #4
 +
| 11000000 00111100
 +
| C0 3C
 +
|-
 +
| #5
 +
| 00000000 00111101
 +
| 00 3D
 +
|-
 +
| #6
 +
| 01000000 00111101
 +
| 40 3D
 +
|-
 +
| #7
 +
| 10000000 00111101
 +
| 80 3D
 +
|-
 +
| #8
 +
| 11000000 00111101
 +
| C0 3D
 +
|-
 +
| #9
 +
| 00000000 00111110
 +
| 00 3E
 +
|-
 +
| #10
 +
| 01000000 00111110
 +
| 40 3E
 +
|-
 +
| #11
 +
| 10000000 00111110
 +
| 80 3E
 +
|-
 +
| #12
 +
| 11000000 00111110
 +
| C0 3E
 +
|-
 +
| #13
 +
| 00000000 00111111
 +
| 00 3F
 +
|-
 +
| #14
 +
| 01000000 00111111
 +
| 40 3F
 +
|-
 +
| #15
 +
| 10000000 00111111
 +
| 80 3F
 +
|-
 +
| #16
 +
| 11000000 00111111
 +
| C0 3F
 
|}
 
|}
  
Line 299: Line 769:
  
 
=== Texture page calculation ===
 
=== Texture page calculation ===
 +
Byte TPage = InputBytes[TexturePage_index] & 0F;
 +
Bitwise TPage byte AND 0F, to delete first 4 bits               
 +
int TPageINT = TPage * 64;
 +
For 16 bit TIM's, the texture page is 64 pixels wide
 +
 +
====Example====
 +
For 24-bit TIM:
 +
0xB2 byte is: 2*48 = 96
 +
Unsure about this one. It could be 42.667 instead of 48. It takes 1.5x the space of 16 bit. I haven't seen a file that uses 24-bit yet.
 +
 +
For 16-bit TIM:
 +
0xB2 byte is: 2*64= 128
  
                string StrByte = InputBytes[TexturePage_index].ToString("X2"); //Gets TPage byte as HEX text
+
For 8-bit TIM:
                StrByte = "0" + StrByte.Substring(1); // Deletes the first char/ 6 bits
+
0xB2 byte is: 2*128 = 256
                Byte TPage = Byte.Parse(StrByte); // Parses result as new byte
 
                int TPageINT = TPage * 128; //For 8 bit TIM's, the texture page is 128 sized
 
  
Example:
+
For 4-bit TIM:
0xB2 byte is:
+
0xB2 byte is: 2*256 = 512
2*128 = 256
 
  
=== TIM Texture width/height resolving ===
+
Each time you half the number of bits you double the amount of data you can store in the same space. So the calculated Texture Page is different.
 
            //Updated - 19.07.2015 by MaKiPL
 
           
 
            Byte[] Stage = File.ReadAllBytes(@"PATH TO .X stage file"); //Load file to memory
 
            Byte[] TIMtexture = { 0x10, 0x00, 0x00, 0x00, 0x09 }; //Initialize 8BPP header
 
            int TIMoffset = ByteSearch(Stage, TIMtexture, 0); //Search in Stage for TIM header
 
            int TIMoffsetCLUTetc = TIM + 18; // Logic pass
 
            UInt16 CLUTsize = BitConverter.ToUInt16(Stage, TIMoffsetCLUTetc); //GET CLUT size to UInt16
 
            TIMoffsetCLUTetc += 2 + (CLUTsize * 512) + 8; // Logic pass - pass CLUT offset, to get to sensitive data
 
            UInt16 szerU = BitConverter.ToUInt16(Stage, TIMoffsetCLUTetc); //Get raw UInt16 value of width from texture data
 
            UInt16 wysoU = BitConverter.ToUInt16(Stage, TIMoffsetCLUTetc + 2); //Get raw UInt16 height value from texture data (real)
 
            int width = szerU * 2; //Calculate real texture width (For 8BPP = width*2)
 
  
 
=== Face order / Translation/ triangulation ===
 
=== Face order / Translation/ triangulation ===
  
'''1.Quad wing order:'''
+
====Quad wing order====
 
   ABCD ---> ABDC
 
   ABCD ---> ABDC
 
   example:
 
   example:
Line 332: Line 799:
 
   f 1 2 4 3
 
   f 1 2 4 3
  
'''2.Quad triangulation face order'''
+
====Quad triangulation face order====
 
   ABDC > ABD
 
   ABDC > ABD
 
         ACD (same for VT)
 
         ACD (same for VT)
Line 340: Line 807:
 
   f 1 2 4
 
   f 1 2 4
 
   f 1 3 4
 
   f 1 3 4
'''
+
====Triangles VT order====
3.Triangles VT order:'''
 
 
   A/T1 > A/T2
 
   A/T1 > A/T2
 
   B/T2 > B/T3
 
   B/T2 > B/T3
 
   C/T3 > C/T1
 
   C/T3 > C/T1
  Example:
+
====Example====
 
   (Also with quad triangulation face order!)
 
   (Also with quad triangulation face order!)
 
   f 1/1 2/2 7/4 6/3
 
   f 1/1 2/2 7/4 6/3

Latest revision as of 21:55, 23 December 2020

By MaKiPL. Thanks for help in research for: shakotay2 (XeNTaX), Halfer, Yagami Light. Complete list of original battle stages by Kaspar01: List of battle stages

Info

.X file is uncompressed 3D stage model with texture embedeed. The file contains unused info (not used by FF8 engine), camera data, movement, translations and geometry data.

Battle stages DOES NOT contain pointers to next sections. All pointers are HARDCODED in FF8.EXE. Click me for battle stage pointer list

Name Usually starting with: Description
PlayStation MIPS assembly E8 FF BD 27 01 00 02 Used only in PS version; skipped in PC
Camera data + movement 02 00 08 00 20 00 Camera animations, movement (pre-keyed)
Model section 06 00 00 00 (See geometry section info) Divides to object (group) and it's sub-objects + vert/triangle grouping
Texture 10 00 00 00 09 (Always) Contains one .TIM texture, 8 BPP

How is this file handled by the engine?

FF8 loads specific file and reads it from hardcoded (written in FF8 code) position (or reads MIPS PlayStation assembly that contains battle stage load code). This hardcoded position points to camera data (in PC). Camera data has size uint16, which is calculated and relative jump is made. Just after camera, a typical section based file handling is made.

Offset Length Description
0 uint32 Number of sections
4 uint32 Objects group #1
8 uint32 Objects group #2
12 uint32 Objects group #3
16 uint32 Objects group #4
20 uint32 Texture [unused in code]
24 uint32 Texture
28 uint32 Relative to EOF

Objects Group

Offset Length Description
0 uint32 Number of sections
4 uint32 Pointer to Settings #1
8 uint32 Pointer to ObjectsList
12 uint32 Pointer to Settings #2
16 uint32 Relative to end_of_group
Objects List
Offset Length Description
0 uint32 Number of objects
Num_of_obj*4 uint32 Relative pointer to segment (one .obj file) [01 00 01 00]
Settings 1
Offset Length Description
0 uint16 Unknown, fixed 2?
2 sint16 (probably, or uint16) Scale (01 is the smallest)
4 4 bytes PADDING
8 4 bytes Nothing?
12 4 bytes Nothing?
16 2 bytes if not FF FF, then model vanishes
18 sint16 X or Z Axis rot origin (Translation)
20 44 bytes Left null for storing some engine data AFTER loaded in-game
64 2 bytes if not FF FF or 00 00, then model vanishes
66 2 bytes Nothing?
68 ...up to end Left null for storing some engine data AFTER loaded in-game

This might look difficult, but it's like multi-pointer file that starts at different location given by executable. See list at the beginning of this wiki to see which stages camera starts at which position (where FF8.exe really starts to read file as BattleStage) [FF8 stores whole file, even with useless data before camera] If this still makes trouble, see this video: TODO

Camera data

Starts at ~0x5d4 (see How the engine handles this file?). [.text:00509820]

Offset Length Description
0 uInt16 Pointers count. Fixed (02 00) [Unused in code]
2 uInt16 Relative jump to CameraSetting. Should be (08 00)
4 uInt16 Relative jump to CameraAnimationCollection. Should be (20 00)
6 uInt16 Camera data size (starting from 0) [unused in code]
8 24 bytes Camera Settings
32 ?? bytes Camera Animation Collection

Camera Setting

Offset Length Description
0 char (based on disassembly) CameraAnimMode?
1 byte? StopEnemyBeforeAnim?
2 Bitfield? (1B) UHM?
3 ??? DefaultEndofCameraPosition_Zoom?
4 ??? ???

Camera Animations Collection

Offset Length Description
0 uint16 NumOfSets
2 *(Set index) uint16 Relative pointer to camera Animation Set
2 *(Set index)+ 2 uint16 Camera EOF
CameraAnimationSetPointer CameraAnimationSet CameraAnimationSet
CameraAnimationSet

Seems to be 8 pointers in each set.

Offset Length Description
0 ushort AnimPointer*2
AnimPointer Varies 144-146-148 bytes usually CameraAnimation

Camera Animation (WIP)

All this is based on best effort knowledge from reversing by Maki. I'm just trying to understand it.

Control Word

Offset Length Description
0 uint16_t (bit varies) Main controller. If 0xFFFFU END

FOV

Control Word & 0b0000'0000'1100'0000U
1 Default
Offset Length Description
None None Start = 0x200
None None End = 0x200
2 Same
Offset Length Description
0 uint16_t Start
2 uint16_t Padding
None None End = Start
2 Different
Offset Length Description
0 uint16_t Start
2 uint16_t Padding
4 uint16_t End

ROLL

Control Word & 0b0000'0011'0000'0000U
0 Unknown
TODO
1 Default
Offset Length Description
None None Start = 0x000
None None End = 0x000
2 Same
Offset Length Description
0 uint16_t Start
None None End = Start
2 Different
Offset Length Description
0 uint16_t Start
2 uint16_t End

LAYOUT

Control Word & 0b0000'0000'0000'0001U
0 Default
You'll loop through the data till the first value is less than 0. pushing back to a variable length container.
Offset Length Description
0 int16_t if < 0 break; could be related to time of frame.
2 16 bytes Animation Frame
1 Other
TODO

Time

Time is calculated from number of frames. You basically set starting position World+lookat and ending position, then mark number of frames to interpolate between them. Every frame is one draw call and it costs 16. Starting time needs to be equal or higher for next animation frame to be read; If next frame==0xFFFF then it's all done.

Animation Frame

Offset Length Description
0 uint16_t is frame durations shot
2 Vertice<uint16_t> (x,y,z) World
8 uint16_t is frame ending shot
10 Vertice<uint16_t> (x,y,z) Look At
Dependency
  CAMERA:
  -Camera Settings
  -Camera Animation Collection (even three collections in a0stg006.x and up to 7 in a0stg101.x!)
     |
     |
      -Camera animation Set (Always 8 camera animations, may be empty (0xFFFF); look for example to a0stg127.x collectionId==1; there are 8 animations, but 7 of them are 0xFFFF (pointers increase by 2))
         |
         |
          - Camera animation


Example

a0stg006.x:
0x5d8 -> 02 00 08 00 20 00

  1. Get EOF-> *(0x5d8 + 8) -> 32
  2. Jump to EOF -> 0x5d8 + 32 = 0x5F8 (This is now Camera Animation Collection)
  3. Get pointer to correct anim collection. In this case we will use AnimCollectionID == 0, so: *(0x5F8 + 0*2 + 2) -> 0x0c
  4. Jump to Anim collection data: 0x5F8 + 0x0c = 0x604 (This is now Camera Animation Set)
  5. Jump to Camera animation by cameraAnimSetID, let's take for example cameraAnimSetID == 7, so: *(0x604 + 7*2) -> 0x25E. Now carefully, jump by multiplying it by 2!
    1. 0x604 + (0x25E * 2) = 0xAC0

Therefore: a0stg006.x Camera animation 7 in camera collection 0 is at 0xAC0

Geometry

Geometry contains groups, that contains Triangles and/or quads poligons. Models always stars at 01 00 01 00, and needs to be after camera data, either it's not model itself, but some other data.

Group

Offset Length Description
0 4 bytes Always 01 00 01 00 / Header of object
4 2 bytes Uint16 / number of vertices.
6 Number of vertices * 6 Vertex data, short X; short Y; short Z;
6 + (Number of vertices *6) (AbsolutePosition MOD 4) + 4 Padding
varies (just after above) 2 bytes uint16 / number of triangles
varies (just after above) 2 bytes uint16 / number of quads
varies (just after above) 4 bytes padding
varies (just after above) number of triangles * 20 Triangle data. If NumOfTriangles = 0, then instead of any triangle data, there's quad data.
varies (just after above) number of quads * 24 Triangle data. If NumOfQuads = 0, then instead of any quad data, there's either next header 01 00 01 00, or end of group.
Triangle
Name Type Description
F1 (A) uint16 A of face indice
F2 (B) uint16 B of face indice
F3 (C) uint16 C of face indice
U1 Byte U of first texture coordinate
V1 Byte V of first texture coordinate
U2 Byte U of second texture coordinate
V2 Byte V of second texture coordinate
CLUT_ID UInt16 (BIT operated)* Index to CLUT_ID operated by BIT (see below)
U3 Byte U of third texture coordinate
V3 Byte V of third texture coordinate
Special (see below / After QUAD table) 8 Byte See below (after QUAD table)
Hide Byte/Bool Bool. Hides or shows texture
Red Byte Texture colourization (Red)
Green Byte Texture colourization (Green)
Blue Byte Texture colourization (Blue)
PSone GPU related Byte PSOne instruction
Quad
Name Type Description
F1 (A) uint16 A of face indice
F2 (B) uint16 B of face indice
F3 (C) uint16 C of face indice
F4 (D) uint16 D of face indice
U1 Byte U of first texture coordinate
V1 Byte V of first texture coordinate
CLUT_ID UInt16 (BIT operated)* Index to CLUT_ID operated by BIT (see below)
U2 Byte U of second texture coordinate
V2 Byte V of second texture coordinate
Special (see below) 8 Byte See below (after table)
Hide Byte/Bool Bool. Hides or shows texture
U3 Byte U of third texture coordinate
V3 Byte V of third texture coordinate
U4 Byte U of fourth texture coordinate
V4 Byte V of fourth texture coordinate
Red Byte Texture colourization (Red)
Green Byte Texture colourization (Green)
Blue Byte Texture colourization (Blue)
PSone GPU related Byte PSOne instruction

Special Byte: This byte, needs to be divided to two bytes. So, for example, a 0xE5 needs to be treated like two bytes: 0x0E 0x05

Singular of char Example Description
First char/ 4 BIT B2 --> 0B Unknown
Second char / 4 BIT B2 --> 02 Texture page number

CLUT ID

The most important bit's are the first two on one byte, and last two on last byte. Example: 00000000 00111100 (00 3C) should be read like this: 00111100 00000000 (3C 00) And the CLUT ID is revealed by watching this four bits: 001111[00 00]000000

Same applies to reading CLUT colors (16bits) where they are: 1b-T, 5b-B, 5b-G, 5b-R (Or RGBT if reordered as shown above)

Complete list (without replacing bit order - as is in HEX editor/memory):

CLUT ID BIT HEX
#1 00000000 00111100 00 3C
#2 01000000 00111100 40 3C
#3 10000000 00111100 80 3C
#4 11000000 00111100 C0 3C
#5 00000000 00111101 00 3D
#6 01000000 00111101 40 3D
#7 10000000 00111101 80 3D
#8 11000000 00111101 C0 3D
#9 00000000 00111110 00 3E
#10 01000000 00111110 40 3E
#11 10000000 00111110 80 3E
#12 11000000 00111110 C0 3E
#13 00000000 00111111 00 3F
#14 01000000 00111111 40 3F
#15 10000000 00111111 80 3F
#16 11000000 00111111 C0 3F

Texture

Contains one TIMs with various size 512x256, 673x256, 768x256 (8BPP).

UV calculation algorithm

Float U = (float)U_Byte / (float)(TIM_Texture_Width * 2) + ((float)Texture_Page/(TIM_Texture_Width * 2));

Texture page calculation

Byte TPage = InputBytes[TexturePage_index] & 0F;

Bitwise TPage byte AND 0F, to delete first 4 bits

int TPageINT = TPage * 64;

For 16 bit TIM's, the texture page is 64 pixels wide

Example

For 24-bit TIM:

0xB2 byte is: 2*48 = 96

Unsure about this one. It could be 42.667 instead of 48. It takes 1.5x the space of 16 bit. I haven't seen a file that uses 24-bit yet.

For 16-bit TIM:

0xB2 byte is: 2*64= 128

For 8-bit TIM:

0xB2 byte is: 2*128 = 256

For 4-bit TIM:

0xB2 byte is: 2*256 = 512

Each time you half the number of bits you double the amount of data you can store in the same space. So the calculated Texture Page is different.

Face order / Translation/ triangulation

Quad wing order

 ABCD ---> ABDC
 example:
  f 1 2 3 4
 to:
  f 1 2 4 3

Quad triangulation face order

 ABDC > ABD
        ACD (same for VT)
 example:
 f 1 2 4 3
 to:
 f 1 2 4
 f 1 3 4

Triangles VT order

 A/T1 > A/T2
 B/T2 > B/T3
 C/T3 > C/T1

Example

 (Also with quad triangulation face order!)
 f 1/1 2/2 7/4 6/3
 to:
 f 1/1 2/2 7/4
 f 1/1 6/3 7/4
 where:
 A=1 B=2 C=6 D=7