it's definitively *not* some float coordinates.
the python script of ZeroSignal is wrong for these fields; it does generate some NaN reading values as float and it does not manage all present data.
coordinates related to the character are (fortunately) not required, the position of the hands and arms are already known (from the character definition), the only possibly needed "spatial information" is the orientation of weapon - but even for that I'm not sure that an info is stored into shape data -- if you equip a spear on left arm (so as a shield) it will be shown perpendicularly to the arm, as a shield.
regarding .itm file, according my understanding the format (of the 3rd block) is
Code:
int32 index of 3DO file
int32 index of DDS file
int32 glow // or int8 glow and 3 unknown bytes
int32 unknown // always 16h
int32 format // 0 -> 4 bytes to follow, 1 -> 20 bytes to follow
int32 unknown // 0x00000000 (0x00000001 in rare cases)
// for format = 1 only
long unknow[4] // may be some float
Edit:
ok, coords may not be necessary unless they are; ZeroSignal interpretation was right - he simply miss variable-length point.
my latest understanding of the .ITM records is:
Code:
int32 index of 3DO file
int32 index of DDS file
int32 glow // 0:blue, 1:red, 2:green, 3:white [1], -1:none
int32 unknown // 0, 11, 12, 21, 22, 24, 26
int32 format // 0 -> 4 bytes to follow, 1 -> 20 bytes to follow
int32 unknown // 0, 1, 3, 4, 5, 6
// for format = 1 only
int32 rgba // color (true 32bits clr with alpha), may be the color of smog during hit?!
float4 posX // coords of what related to what ?...
float4 posY // coords of what related to what ?...
float4 posZ // coords of what related to what ?...
[1] as per ProfNerwosol, Guide to Shaiya Configuration
the color field has to be investigated with higly visible color (full red ...), the checked files contain only some gray with very small differences between them; it so wasn't possible to determine what item, part of item, event, anim is colored, if any.