Writing a tiny unpacker for a mystery .pak format

Magic bytes, offset tables, and a weekend in a hex editor — the first step of the OreGairu English Patch.

Looking at the header

Every archive starts with the same four bytes, followed by a file count and the offset of a table of contents.

import struct

def read_header(data: bytes):
    magic, count, toc = struct.unpack_from("<4sII", data, 0)
    assert magic == b"PAK\0", "not a pak file"
    return count, toc

How the pieces fit

Repacking

Repacking is the same walk in reverse. The only rule: keep each file aligned to 16 bytes, or the game refuses to load it. The cleaned-up version of this script became paktool.

• Read the header

• List every entry

• Extract files

• Repack with alignment