Extractor for unity3d files from Higurashi Mei

HiguMeiExtractor.py

import unitypack
import unitypack.utils
import collections
import sys
import os
from unitypack.engine.texture import TextureFormat
from PIL import ImageOps

ASTC_MAP = {
	TextureFormat.ASTC_RGB_4x4:    (0x93B0, 0x1907),
	TextureFormat.ASTC_RGB_5x5:    (0x93B2, 0x1907),
	TextureFormat.ASTC_RGB_6x6:    (0x93B4, 0x1907),
	TextureFormat.ASTC_RGB_8x8:    (0x93B7, 0x1907),
	TextureFormat.ASTC_RGB_10x10:  (0x93BB, 0x1907),
	TextureFormat.ASTC_RGB_12x12:  (0x93BD, 0x1907),
	TextureFormat.ASTC_RGBA_4x4:   (0x93B0, 0x1908),
	TextureFormat.ASTC_RGBA_5x5:   (0x93B2, 0x1908),
	TextureFormat.ASTC_RGBA_6x6:   (0x93B4, 0x1908),
	TextureFormat.ASTC_RGBA_8x8:   (0x93B7, 0x1908),
	TextureFormat.ASTC_RGBA_10x10: (0x93BB, 0x1908),
	TextureFormat.ASTC_RGBA_12x12: (0x93BD, 0x1908),
	TextureFormat.ETC2_RGBA8:      (0x9278, 0x1908)
}

PVRTC_MAP = {
	TextureFormat.PVRTC_RGB2:  (0xC, 0x10000, 2),
	TextureFormat.PVRTC_RGBA2: (0x18, 0x18000, 2),
	TextureFormat.PVRTC_RGB4:  (0xD, 0x10000, 4),
	TextureFormat.PVRTC_RGBA4: (0x19, 0x18000, 4)
}

def processTex2D(info, outpath):
	if info.format in ASTC_MAP:
		with open(outpath + ".ktx", "wb") as outImage:
			outImage.write(bytes([0xAB, 0x4B, 0x54, 0x58, 0x20, 0x31, 0x31, 0xBB, 0x0D, 0x0A, 0x1A, 0x0A])) # identifier
			outImage.write(0x04030201.to_bytes(4, byteorder="little")) # Endianness
			outImage.write(b"\0" * 12) # glType, glTypeSize, glFormat
			outImage.write(ASTC_MAP[info.format][0].to_bytes(4, byteorder="little")) # glInternalFormat
			outImage.write(ASTC_MAP[info.format][1].to_bytes(4, byteorder="little")) # glBaseInternalFormat
			outImage.write(info.width.to_bytes(4, byteorder="little")) # pixelWidth
			outImage.write(info.height.to_bytes(4, byteorder="little")) # pixelHeight
			outImage.write(b"\0" * 8) # pixelDepth, numberOfArrayElements
			outImage.write(0x1.to_bytes(4, byteorder="little")) # numberOfFaces
			outImage.write(0x1.to_bytes(4, byteorder="little")) # numberOfMipmapLevels
			keyAndValue = b"KTXorientation\0S=r,T=u\0" # flip upside down
			byteCount = len(keyAndValue) + (3 - ((len(keyAndValue) + 3) % 4))
			outImage.write((4 + byteCount).to_bytes(4, byteorder="little")) # bytesOfKeyValueData
			outImage.write(len(keyAndValue).to_bytes(4, byteorder="little")) # keyAndValueByteSize
			outImage.write(keyAndValue + b"\0" * (byteCount - len(keyAndValue))) # keyAndValue, padding
			outImage.write((info.width * info.height).to_bytes(4, byteorder="little")) # imageSize
			outImage.write(info.image_data) # image data
		return
	if info.format in PVRTC_MAP:
		with open(outpath + ".pvr", "wb") as outImage:
			# outImage.write(0x03525650.to_bytes(4, byteorder="little")) # Version
			# outImage.write(b"\0" * 4) # Flags
			# outImage.write(PVRTC_MAP[info.format].to_bytes(8, byteorder="little")) # Pixel Format
			# outImage.write(0x1.to_bytes(4, byteorder="little")) # Colour Space (sRGB)
			# outImage.write(b"\0" * 4) # Channel Type (Unsigned Byte Normalized)
			# outImage.write(info.height.to_bytes(4, byteorder="little")) # Height
			# outImage.write(info.width.to_bytes(4, byteorder="little")) # Width
			# outImage.write(0x1.to_bytes(4, byteorder="little")) # Depth
			# outImage.write(0x1.to_bytes(4, byteorder="little")) # Num Surfaces
			# outImage.write(0x1.to_bytes(4, byteorder="little")) # Num Faces
			# outImage.write(0x1.to_bytes(4, byteorder="little")) # Mipmap Count
			# outImage.write(b"\0" * 4) # Metadata Size
			# outImage.write(info.image_data) # image data

			outImage.write((52).to_bytes(4, byteorder="little")) # Header Size
			outImage.write(info.height.to_bytes(4, byteorder="little")) # Height
			outImage.write(info.width.to_bytes(4, byteorder="little")) # Width
			outImage.write(0x1.to_bytes(4, byteorder="little")) # Mipmap Count
			outImage.write(PVRTC_MAP[info.format][0].to_bytes(1, byteorder="little")) # Pixel Format
			outImage.write(PVRTC_MAP[info.format][1].to_bytes(3, byteorder="big")) # Flags
			outImage.write(len(info.image_data).to_bytes(4, byteorder="little")) # Surface Size
			outImage.write(PVRTC_MAP[info.format][2].to_bytes(4, byteorder="little")) # Bits per Pixel
			outImage.write(b"\xff\0\0\0") # Red Mask
			outImage.write(b"\0\xff\0\0") # Green Mask
			outImage.write(b"\0\0\xff\0") # Blue Mask
			outImage.write(b"\0\0\0\xff") # Alpha Mask
			outImage.write(b"PVR!") # PVR Identifier
			outImage.write(0x1.to_bytes(4, byteorder="little")) # Number of Surfaces
			outImage.write(info.image_data) # image data
		return
	try:
		image = info.image
	except NotImplementedError as error:
		print("Failed to decode " + outpath + " with unsupported format " + str(info.format))
		return
	if not image:
		return
	img = ImageOps.flip(image)
	with open(outpath + ".png", "wb") as outImage:
		img.save(outImage, format="png")

def extract_asset(path, asset, print_ignored = True):
	if asset.type in ("AssetBundle", "MonoScript", "AnimationClip", "GameObject", "Transform", "Mesh", "Sprite", "Avatar", "CharacterImasMotionAsset"):
		return False

	asset = asset.read()

	outpath = os.path.join(sys.argv[1], path)
	if not os.path.exists(os.path.dirname(outpath)):
		os.makedirs(os.path.dirname(outpath))

	if not os.path.splitext(outpath)[1]:
		outpath += ".bin"

	if isinstance(asset, unitypack.engine.Texture2D):
		processTex2D(asset, os.path.splitext(outpath)[0])
	elif isinstance(asset, unitypack.engine.TextAsset):
		with open(outpath, "wb") as outfile:
			if isinstance(asset.bytes, str):
				outfile.write(asset.bytes.encode("utf-8"))
			else:
				outfile.write(asset.bytes)
	elif isinstance(asset, unitypack.engine.AudioClip):
		samples = unitypack.utils.extract_audioclip_samples(asset)
		for name, sample in samples.items():
			with open(os.path.join(os.path.dirname(outpath), name), "wb") as outfile:
				outfile.write(sample)
	else:
		if print_ignored:
			print(infile)
			print(type(asset))
			print(entry[1]["asset"].object.type)
		return False
	return True

for infile in sys.argv[2:]:
	with open(infile, "rb") as file:
		bundle = unitypack.load(file)
		extracted = False
		path = ""
		for e in bundle.assets[0].objects.values():
			if e.type != "AssetBundle": continue
			for entry in e.read()["m_Container"]:
				path = entry[0]
				if path[-6:] == ".bytes":
					path = path[:-6]

				if entry[1]["asset"] is None:
					continue
				target = entry[1]["asset"]

				if extract_asset(path, target.object):
					extracted = True

		if not extracted and path:
			# We don't extract 3D models, but try to extract their textures
			i = 0
			for item in bundle.assets[0].objects.values():
				if item.type == "Texture2D" and item.read().name:
					name = item.read().name

					if os.path.basename(os.path.splitext(path)[0]).lower() == name.lower():
						name = os.path.join(os.path.dirname(path), name)
					else:
						name = os.path.splitext(path)[0] + "_" + name
					extract_asset(name, item, print_ignored=False)
				elif item.type == "TextAsset" and item.read().name:
					name = os.path.splitext(path)[0] + "_" + item.read().name + ".txt"
					extract_asset(name, item)
				elif extract_asset(os.path.splitext(path)[0] + "_" + str(i), item, print_ignored=False):
					i += 1


				

Also is there a way to change where the decompression goes? I want it in the D Drive however it seems like my computer is using the C drive to decrypt it. This is what I have been getting

Traceback (most recent call last):
File "D:\Art\Apps\Higurashi Mei\Higurashi Mei Extractor\9188559caf704f400d63db8740898a77-332733ee7d856609781fe86dedec6b886cca9802\HiguMeiExtractor.py", line 132, in
bundle = unitypack.load(file)
File "C:\Users\USER\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\unitypack_init_.py", line 12, in load
return env.load(file)
File "C:\Users\USER\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\unitypack\environment.py", line 27, in load
ret.load(file)
File "C:\Users\USER\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\unitypack\assetbundle.py", line 49, in load
self.load_unityfs(buf)
File "C:\Users\USER\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\unitypack\assetbundle.py", line 106, in load_unityfs
data = self.read_compressed_data(buf, compression)
File "C:\Users\USER\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\unitypack\assetbundle.py", line 92, in read_compressed_data
return lz4_decompress(data, self.uiblock_size)
File "C:\Users\USER\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\unitypack\utils.py", line 11, in lz4_decompress
return decompress(data, size)
_block.LZ4BlockError: Decompression failed: corrupt input or insufficient space in destination buffer. Error code: 16

is input1.unity3d and input2.unity3d placeholders? Or do I just copy what you have here and put that into the command prompt and hit enter? And so I need to do anything to the apk of the game such as unzip it?

All of those are placeholders. output-directory is the place it will output, so replace that with D:\whateverblahblah

Looking back on the discord history from this, it looks like the unity3d files were a part of an encrypted archive. If you join the 07th-mod discord then this link should get you to the chat history from that.

Otherwise, the decryption tool is here: https://github.com/TellowKrinkle/SxamTools. If you currently have a list.bin, data.bin, and hash.bin, this is the tool to use. They may have changed the encryption key since then, since that was two years ago. If so, it should fail with a hash failure, and you'll have to get the new key by decompiling the game. If you don't want to set up a swift compiler, someone also made a C++ version: https://gist.github.com/viliml/e70ee7c9b14c7591ee3bd14809e2851c

If you currently have the game apk, that's not enough unfortunately. The game downloads its assets after completing the prologue, so the apk won't have them. I wasn't the one who got the original asset file, but I'd guess it involved using an Android file manager to grab the file out of the application's resources after running it. Might require a rooted phone, idk.

is input1.unity3d and input2.unity3d placeholders? Or do I just copy what you have here and put that into the command prompt and hit enter? And so I need to do anything to the apk of the game such as unzip it?

All of those are placeholders. output-directory is the place it will output, so replace that with D:\whateverblahblah

Looking back on the discord history from this, it looks like the unity3d files were a part of an encrypted archive. If you join the 07th-mod discord then this link should get you to the chat history from that.

Otherwise, the decryption tool is here: https://github.com/TellowKrinkle/SxamTools. If you currently have a list.bin, data.bin, and hash.bin, this is the tool to use. They may have changed the encryption key since then, since that was two years ago. If so, it should fail with a hash failure, and you'll have to get the new key by decompiling the game. If you don't want to set up a swift compiler, someone also made a C++ version: https://gist.github.com/viliml/e70ee7c9b14c7591ee3bd14809e2851c

If you currently have the game apk, that's not enough unfortunately. The game downloads its assets after completing the prologue, so the apk won't have them. I wasn't the one who got the original asset file, but I'd guess it involved using an Android file manager to grab the file out of the application's resources after running it. Might require a rooted phone, idk.

Do you know how to get the encryption key by decompiling , using which tools thank you！

I used ghidra with Il2CppDumper

It was in Sxam.Setting_TypeInfo->static_fields->DecodeKey, which was initialized in Sxam.Setting$$ctor and passed to Sxam.Crypto$$Decode