Sisters - Saigo natsu no hi - can't find in-game text

[Tai]

Hi

Seeing as I have loads of free time i thought it'd be nice to try myself at hacking vn's. I started with SISTERS, because my friend told me he'd translate it if I suceed.

And so here's my problem:

After writing an unpacker and unpacking the content from game archives (.isa) i got loads of .png's (obviously) and some game script .ism files, in which i expeted to find kanji that i could extract, pass to my TL, and paste back translated. Well i know it wouldn't be that easy, but you catch my drift. Sadly those files don't seem to contain any interesting data at all, nor they seem to be encrypted.

http://speedy.sh/ZtjW8/data0.zip <- i collected them all here

Any idea if I can do something with them?

[erengy]

I took a quick look at a couple of .ISM files, and the file structure seems to be something along the lines of:

Spoiler

char* magic ("ISM SCRIPT")
int32 file_size

int32 header_end_offset

(Here's a list of something. May be triplets of int32s?)

int8 length (of the following string)
char* file_name (e.g. "CHAP1_A%1")
int32 (length of instructions?)
int32
int32

int8 length (of the following string)
char* function_name (always "CountParagraph%0"?)
int32 function_count
int32 function_count (?)
int32

int8 length (of the following string)
char* function_name ("%n" probably defines how many parameters that function takes)

(Instructions begin...)

From here, you can either find how each instruction is laid out and code a full-fledged script converter, or just focus on text-related ones to extract character names, dialogs and stuff. Since you seem to have lots of time on your hands, I'd suggest the former.

[Tai]

Yeah, but, doesn't it seem wierd there's no actual text in them? How will I know where to put translated text back to?

[erengy]

Since we can't directly see meaningful words, my guess is that they're encoded in some way. Also the parts that resemble strings are rather short even for a game like this one.

If I were you, I would:

1. Make sure there are no other files (including the game executable) that could contain the text.
2. Map each function name to their corresponding instruction.
3. List the input of text-related functions.
4. Compare them with what you see in game to figure out how they're encoded. Looking out for single-character lines ("？") or repeated characters ("………") might help.

[Tai]

Yeah, turns out that those .ism are encrypted and contain everything. I've set some hwbp's on memory and pinpointed the decrypting part of code.

Each byte is first negated and after that XOR'ed. There are multiple XOR keys even for one .ism