This topic was archived here:

https://web.archive.org/web/20230000000000fw_/https://www.zenhax.com/viewtopic.php?t=14081

TwoLeggedRat, posted Sat Sep 05, 2020 6:25 am (58552)

---

The data.bin file contain the assets for images, sprites, sound, audio, pretty much everything, and using any Unity programs will not open them up.

APK Data.bin (64.9 MB)

Full in-game downloaded Data.bin (776 MB)

I'm providing both the APK and the fully downloaded data.bin from the game's own downloader.

While Data.bin is the presumed to be the main archive containing the assets, both downloads are also supplied with the respective hash.bin, inst.bin and list.bin files presumably necessary to figure this out. (Thanks to soneek for making me aware I should point this out.)

Please help me where I can't help myself, and thank you very much for trying to help!

soneek, posted Mon Sep 07, 2020 11:04 pm (58571)

---

It might be worth including list.bin, and hash.bin as well. inst.bin didn't seem useful.

Edit: They are already included. I already have the files on my end, so I hadn't checked your upload. Sorry about assuming.

TwoLeggedRat, posted Tue Sep 08, 2020 5:53 am (58574)

---

soneek wrote:

It might be worth including list.bin, and hash.bin as well. inst.bin didn't seem useful.

Both downloads contain hash.bin, inst.bin and list.bin along with the Data.bin (that differs between the links as described.)

soneek, posted Tue Sep 08, 2020 5:46 pm (58577)

---

I was able to get the file list from RAM, which might have parts of list.bin decrypted. It corresponds to the large data.bin in terms of offsets and files size, starting around offset 0xd5d8 in the attached file. The first portion of the list has folder names, and the actual file blocks have file names, folder, offset, size, and some other info. I will write a BMS script later based off of it for extraction at least if I get some extra time. As TwoLeggedRat mention, the encryption is the tough part.

TwoLeggedRat, posted Tue Sep 08, 2020 6:32 pm (58579)

---

soneek wrote:

I was able to get the file list from RAM, which might have parts of list.bin decrypted. It corresponds to the large data.bin in terms of offsets and files size, starting around offset 0xd5d8 in the attached file. The first portion of the list has folder names, and the actual file blocks have file names, folder, offset, size, and some other info. I will write a BMS script later based off of it for extraction at least if I get some extra time. As TwoLeggedRat mention, the encryption is the tough part.

It's exciting to see progress! No rush, pressure, or insistence, I'm happy just to see things having gotten this far, so just for this much alone, I REALLY thank you, as I didn't think I'd even get this far.

soneek, posted Wed Sep 09, 2020 3:30 am (58588)

---

So, it looks like the list.bin I got from RAM is incomplete, but here's what I have figured out so far.

Folder table

Code:

long folderCount

for each folder

Code:

long fileCount
string folderName

After the folder table is the file table

Code:

long totalFileCount

For each file entry

Code:

string fileName <- this is the full path without the leading root level folder
short folderId <- This is the array index for the folders read in the folder table
short folderFileNumber <- This is a counter for the file number in the folder
short unk1
long offset <- offset in data.bin
long unk2
long size
longlong someKey <- Maybe something to do with decryption?

I attached a cleaner list.bin from RAM, though it's missing a lot of file entries on the tail end.

TwoLeggedRat, posted Wed Sep 09, 2020 8:04 am (58593)

---

soneek wrote:

So, it looks like the list.bin I got from RAM is incomplete, but here's what I have figured out so far.

Folder table

Code:

long folderCount

for each folder

Code:

long fileCount
string folderName

After the folder table is the file table

Code:

long totalFileCount

For each file entry

Code:

string fileName short folderId short folderFileNumber short unk1
long offset long unk2
long size
longlong someKey 

I attached a cleaner list.bin from RAM, though it's missing a lot of file entries on the tail end.

Getting closer, but it would figure there might be a key necessary as that's how a lot of compressed and encrypted files are now..

soneek, posted Wed Sep 09, 2020 10:03 pm (58607)

---

So, it seems that every file in data.bin is a .unity3d file. I was able to fetch 29 of them from RAM (all 0x1000 bytes or less), and I've attached them along with their encrypted .bin version.

It looks like all of the decrypted files should look like this for the first 0x21 bytes.

Code:

55 6E 69 74 79 46 53 00 00 00 00 06 35 2E 78 2E 78 00 32 30 31 39 2E 34 2E 38 66 31 00 00 00 00 00

The red rectangle is the .unity3d (or .bin) file size in big endian.

TwoLeggedRat, posted Wed Sep 09, 2020 11:22 pm (58608)

---

soneek wrote:

So, it seems that every file in data.bin is a .unity3d file. I was able to fetch 29 of them from RAM (all 0x1000 bytes or less), and I've attached them along with their encrypted .bin version.

It looks like all of the decrypted files should look like this for the first 0x21 bytes.

Code:

55 6E 69 74 79 46 53 00 00 00 00 06 35 2E 78 2E 78 00 32 30 31 39 2E 34 2E 38 66 31 00 00 00 00 00

The red rectangle is the .unity3d (or .bin) file size in big endian.

I'm surprised they're all in the form of unity3d instead of just direct assets, but the good part about this though is that at least they're viewable in something like Asset Studio which makes me VERY much so relieved and calmed for any more future files. I almost forgot how it worked so trying to open them in it confused me but the samples are TextAssets and MonoBehaviour.

I personally don't know what I could do with either of those or how but either way, but this is VERY great news! I only hold out that if any character/sprite imagery is accessible, it's not gonna have any extra issues like no transparency or anything, because that'd be just my luck, all the actual image assets aren't transparent, or they're setup weirdly.

TwoLeggedRat, posted Mon Nov 16, 2020 10:45 pm (60343)

---

Some time later, it seems like someone made a "Sxam" extractor, but I haven't the slightest idea on how it could be put to any use. It seems (relatively?) simple.

bonder, posted Sun Dec 06, 2020 9:22 pm (60944)

---

Also interested in finding out how to extract this. Thank you for the help so far!

aaa801, posted Mon Dec 07, 2020 12:55 am (60948)

---

Ported it over to go and added releases here https://github.com/clienthax/higurashim ... /tag/0.0.1

bonder, posted Wed Dec 09, 2020 2:32 am (60994)

---

aaa801 wrote:

Ported it over to go and added releases here https://github.com/clienthax/higurashim ... /tag/0.0.1

This worked perfectly, thank you so much!

TwoLeggedRat, posted Thu Dec 10, 2020 8:47 am (61028)

---

aaa801 wrote:

Ported it over to go and added releases here https://github.com/clienthax/higurashim ... /tag/0.0.1

Thank you for making this, but I'm not quite sure how to use it, as opening it (administrator or not) just flashes a very quick window and then closes, and dragging anything inside the Sxam folder (or anywhere else resembling it) onto it does the same thing.

EDIT: Nevermind, figured it out! I forget that you usually need to use things like this that don't have UI's developed manually via command prompts! With this said, thank you SO so very much!

bonder, posted Thu Dec 24, 2020 10:03 pm (61273)

---

The extractor works perfectly for most files, however I've run into an issue with these .bytes files. There are quite a few that can be read in notepad , but I believe these have images inside and I'm not sure how to unpack them. I've tried the usual unity-based extrators and have come up with nothing for them. Any help would be appreciated!

aaa801, posted Fri Dec 25, 2020 7:24 pm (61281)

---

Hmm, what one for example?

bonder, posted Fri Dec 25, 2020 10:18 pm (61283)

---

Oh gosh I completely forgot to attach the files! Here's a small sample.

TwoLeggedRat, posted Sun Dec 27, 2020 9:11 am (61310)

---

I was wondering about those myself, I think those might just be how they're packaged inside the data bins, how to convert them and or otherwise extract images from those, I definitely wouldn't have the slightest clue personally.

aaa801, posted Sun Dec 27, 2020 11:22 am (61312)

---

It looks like those havn't been extracted using the tool I linked?,
They should be UnityFS files.

bonder, posted Sun Dec 27, 2020 2:33 pm (61316)

---

Those are the files after they've been extracted and converted via AssetStudio. Here are the raw files as they export after using your tool. I'm still not sure what to do with them

aaa801, posted Sun Dec 27, 2020 2:41 pm (61317)

---

Hm not too sure, looks like someone ran across the same format before https://zenhax.com/viewtopic.php?t=7320

bonder, posted Sun Dec 27, 2020 9:11 pm (61335)

---

I did find a decrypter for that game here: https://github.com/28598519a/DecStarlyGirls
Both games were developed by Smileaxe, but I've no idea how to use this program or if it'd even work with Mei's files.

MLSpence, posted Thu Jan 14, 2021 9:58 pm (61668)

---

Hey so I have been spending my entire day trying to figure out how to rip the files from this game and can't seem to get anywhere, can someone give me a step-by-step guide on how to do this? I've been trying a lot of the stuff suggested through this thread before and can't seem to get anything working, so I believe I'm doing something completely wrong.

manpaint, posted Sun Jan 17, 2021 1:41 am (61707)

---

I have been trying to extract the data from the Sx1.05 files since two day now. I have not yet managed to do anything but I have recorded my finding. It may be useful to someone (see attachement).

TLDR of my findings:

-Sx1.05 are likely archive types. In the case of adventure background, each files seems to contains only one (conjecture).
-Sx1.24 files apparently exist according to the RAM.
-Advbg_001.bytes seems to contains a Advbg_001.bin file according to RAM.
-Ninjaripper and GLintercept do not work with Nox Emulator.
-Starly Girls' SX files have a clear file (archive-like) structure wheras Mei does not.
-DecStarlyGirls does something to the files, but I think its designed to exact .unity3d files from a master file instead of decrypting SX files.
-AdvBg_011 is apparently this background ( https://cdn.discordapp.com/attachments/384427950478589972/800177194508222505/AdvBg_011.png )

-SX files hex header appears to be:

Code:

53 78 31 2E 30 35 00 03 XX XX XX XX XX XX XX XX 00 04 00 04 00 00 08 00 08 00 00 00 
00 02 FF FF FF FF F8 FF FF FF F8 FF FF FF F8 FF FF FF F8 FF FF FF F8 FF FF FF F8 FF

Research notes.txt also contains possible relasted data that I extracted from Nox emulator's RAM.

Edit: SxamTools was updated to support SX files but its written in swift and therefore don't work well with with Windows. If someone could try to port the code it would be great. https://github.com/tellowkrinkle/SxamTools

According to tellowkrinkle, Sx1.05 files are compressed etc1 4bpp texture data, but they've split images into two pieces, each with only two color channels (blue is approximately zero in both)