here is a quick Noesis python script to open your samples until aspadm releases his complete solution. supports astc 6x6, top level mip only

install Adobe Air https://www.adobe.com/products/air.html then install this ATF plugin https://vamapaull.com/atf-to-png-converter-tool/ rename the extension of your samples from dxt to atf then open them with the converter and they will convert to png immediately and write to a folder named "PNG" in...

you must post some samples for examination.
perhaps the samples you tried aren't yet supported by his tex2dds script:
viewtopic.php?f=10&t=11900&p=112757&hil ... +3#p101663

shakotay2 wrote: ↑

Wed Sep 25, 2019 9:28 am

I wonder how you get this script to run? Which Noesis version do you use?
I always get a Type error (Noesis debug output):

i think is because his script is posted with incomplete last line:
return should be return 1

2. my second issue is in the .V2 script, when trying to bind the vertices i dont get the correct data, verts are in half floats , but noesis offers only BYTE, SHORT, FLOAT RPGEODATA_HALFFLOAT seems to still work fine when i tried in this basic script with hardcoded values. :D from inc_noesis import...

try this

your samples are headerless, you must guess width, height and format
and set that in your script accordingly, or find a better way to extract the
textures along with their header info so no need to guess.

what part of solution did you find, any good source code?
it looks like each mip is lzma compressed though.

there is some tools here to work on atf files. :D https://github.com/hajimeku/atf-test-tool or more specifically here: https://github.com/hajimeku/atf-test-tool/tree/master/atftools/batchConvert2ATF/atftool ATFViewer.exe is a GUI tool that will open your 2 samples for viewing and some info, and atf2...

for some reason there is an extra 2 bytes in the decompressed rx3,
i don't know why, all i can say is upload more non working samples and a
solution might be found at some point down the road.

eh i don't really look into model formats much anymore, no time, too busy with texture formats.
all of my current research on the rax format is here though if you care to have a read through:
viewtopic.php?f=16&t=13867&hilit=rax&start=60#p126467

EDXZ23 wrote: ↑

Tue Aug 06, 2019 1:51 am

When you say they cant be read in a single array, what does that mean?

i mean the indices are stored together but i think they are split into material groups,
because when you try to render the whole array they look mangled.
there might be start and stop offsets stored in the file.

the starting offset for vertices looks like it should be 0x3c70
and the starting offset for face indices should be 0x26460
but i don't think you can read them in a single array.
the file structure looks a lot like the rax files from SWBF3.

okay updated the rx3 to dds bms script here for chunlzma support.
viewtopic.php?f=10&t=18916&p=149279#p149279

this Noesis python script will open your 2 samples. supports morton swizzled dxt5.
you must give the samples the tex extension.