XeNTaX

episoder wrote:okay. raw stuff. did the extraction in bms. it outputs dx10 ddses. was easier to deal with bc3+ files. works with batch folder and reimport, but ofc you need a dx10 dds, the same format and sizes. you need the full texture data chunk with mipmaps. in that case they are needed - like most models will need them - you gotta generate them when saving the modified texture.

now... i'm still waiting for the test here. dunno if the de-archiver can reimport this correct now.

edit: probably luck the compressed chunk fit in there, but we got a skintone bra.

Hey, nice work!

I am trying to do the same you did and came pretty far but I fail at using the right settings for saving my .dds because the textures look broken when put in game. I am using Photoshop with the Intel .dds plugin. But it does not have the bc3+ compression you are talking about as a option. Could you tell me what program you used to edited your .dds? Thanks!

episoder wrote:okay. raw stuff. did the extraction in bms. it outputs dx10 ddses. was easier to deal with bc3+ files. works with batch folder and reimport, but ofc you need a dx10 dds, the same format and sizes. you need the full texture data chunk with mipmaps. in that case they are needed - like most models will need them - you gotta generate them when saving the modified texture.

now... i'm still waiting for the test here. dunno if the de-archiver can reimport this correct now.

edit: probably luck the compressed chunk fit in there, but we got a skintone bra.

Excuse me, how to convert from dds to btex

Just copy pure pixels from one file to another, without headers etc. If modified file has the same format as the original one, then pixels will fit without any problems.

...or use QuickBMS.

Robin wrote:I have extracted btex file from fntbin file

How you extract it? I want to replace russian font with something since it's awful.

Open fntbin in hexeditor and search "SEDBbtex" - this is start of the btex file. Except one file, where are stored two BTEX files (I don't remember which one), BTEX content is till the end of fntbin file.

But you have to set correct kerning etc. This info is stored in fntbin, starting from 20205h (font_event.dx11.fntbin) and there are 9 bytes for every character:
1) offset from left of the whole texture, 2 bytes
2) offset from top, 2 bytes
3) texture start, from left, 1 byte (can be negative: 255 means -1) - kerning on the left side of character
4) texture start, from top, 1 byte (vertical placement of character)
5) texture width, 1 byte (how many pixels game have to read starting from 1))
6) texture height, 1 byte
7) whole block of character width (kerning on the right side of character), 1 byte. Example: If you have character with start 2 pixels after beginning of the block - 3) is set to "2", then 7) should be 2 (kerning from left) + texture width + 2 (kerning from right).

Try change something and check how it looks in game. It only looks difficult

I hope i can avoid kerning edits. Will try to make same sized letters just align and smooth them out.
Look at that. I think the person who made it was dead drunk or was a bear.

So I made it, but with all the ajustments needed to make letters fit old placements it doesen't look any better.

What is wrong with these letters? I see problems with kerning in "т", "с" and "у", but don't know cyrillic alphabet too much to see another issues. Also, game is scaling these fonts, on my computer it would be looking even worst

Inconsistent kerning, some letters looks smaller than others like "д" in одному, asymmetric letters like "т" has longer stroke on the right etc.
I noticed that bytes in XY are reversed for some reason. For example 265 written as 09 01 instead 01 09. Took me an hour to determine where 10th byte coming from.
Here I photoshoped calibri over original. I want it to be like this.
http://www.framecompare.com/screenshotc ... n/J2C1NNNU

Yeah, so kerning has to be modified. SE probably used Pro versions of some CJK fonts and just copied kerning to the game.

As for bytes, they are always reversed, as they are arranged in little-endian format.