v2.0.2.1

• "Fixed" -pa not taking .ini into account. Though it only works when you drop .ini on the splitter, not .tga. For .tga it tries to find .tga.ini.preload.
  Btw if your antivirus tries to burn this with fire and virustotal shows like 30+ positives, blame Microsoft, not me. It's their statically linked VC runtime that causes this. I've checked.

v2.0.2.0

• Support for .ini (int .preload in text format from older games).
• Corrected the middle&offsets calculations (AGAIN ._ .). Now should work correctly for episodes and for CIU (2.0.1.4- only worked correctly with CIU).
• New packing debug option --dbg-show-transparency that tints transparent pixels purple. Helps locating stray pixels with 1 alpha which are preventing the atlas from being smaller.
• --dbg-frame: now marks the middle pixel with a darker colour on each side of the frame (local frame when packing, global frame when exporting).
• --dbg-frame, --dbg-middle: now work for packing as well, for some reason.

If you are using this with Krita, remember that it sets the 18th byte in the Targa header (image descriptor) to 0x28 instead of 0x08 which changes the reading order of pixels. What you see correctly is actually flipped vertically, and there's no way to ask Krita not to do this. UVEPS keeps the byte the way it was in the original header, so if you see that the texture is fine but in-game it's flipped and all of the frames are messed up, just flip it vertically again.
Same goes for imagemagick. For example, I used this script to convert all PNGs in the folder (exported from Blender 4.x) to greyscale TGA files.

set MAGICK=C:\Program Files\ImageMagick-7.1.1-Q16-HDRI\magick.exe
set DND_DIR=%~dp1

@rem -colorspace Gray -level 0%%,65%%
for %%f in ("%DND_DIR%*.png") do (
    if "%%~xf"==".png" (
        "%MAGICK%" convert -flip -colorspace Gray -level 0%%,65%% -resize 75x75 "%%~ff" "%%~dnpf.tga"
    )
)
pause

ImageMagick sets 0x20 bit, so for episodes (which don't) I have to add -flip. When you open such files they will appear flipped, but for UVEPS they are fine. Then you just pack them with --pack int and the resulting atlas will also have this bit set. You can manually overwrite it back to 0x08 using a hex editor to see what's the actual result. Obviously, if you are working with CIU, the -flip flag would actually not be needed.

v2.0.1.4

• Fixed alpha channel check for transparent blitting.

Yet another another update with a colour bleeding fix. Please don't use 2.0.1.3, it just doesn't work correctly.

v2.0.1.3

• Added a colour bleeding padding option. See "--colour-padding".
• Added --greyscale option to convert 32 bit frames to 8 bit atlas.
• Added a resource file with general information.

This release aims at fixing bilinear filtering seams that appear due to transparent pixels having a plain white/black colour in the image editor.

v2.0.1.2

• Fixed an issue where free rect reference became outdated after pushing back at least one new rect.
• Changed one particular C-style cast to C++ style cast.
• Removed useless code comments.
  LICENSE.txt