find and extract MP3 files from a binary file

exmp3.py

#!/usr/bin/env python3
from math import floor
import struct

mp3_bit_rates = {
    0b0001:  32000,
    0b0010:  40000,
    0b0011:  48000,
    0b0100:  56000,
    0b0101:  64000,
    0b0110:  80000,
    0b0111:  96000,
    0b1000: 112000,
    0b1001: 128000,
    0b1010: 160000,
    0b1011: 192000,
    0b1100: 224000,
    0b1101: 256000,
    0b1110: 320000,
}

mp3_sampling_rates = {
    0b00: 44100,
    0b01: 48000,
    0b10: 32000,
}

def find_mp3(stream, prefix):
    counter = 0
    is_mp3 = False
    last_mp3_bytes = b""
    my_bytes = b"\0" + stream.read(3)

    while True:
        if not is_mp3:
            if len(last_mp3_bytes) > 0:
                file_name = "{0}{1:04}.mp3".format(prefix, counter)
                counter += 1
                with open(file_name, "wb") as f:
                    f.write(last_mp3_bytes)
            last_mp3_bytes = b""

        is_mp3 = False

        my_bytes = my_bytes[1:] + stream.read(1)
        if len(my_bytes) != 4:
            # it's over
            return

        (header_number,) = struct.unpack(">I", my_bytes)

        # frame sync: 0b11111111 111xxxxx xxxxxxxx xxxxxxxx
        if (header_number & 0xFFE00000) != 0xFFE00000:
            # nope
            continue

        # MPEG version: 0bxxxxxxxx xxxVVxxx xxxxxxxx xxxxxxxx
        mpeg_version = (header_number & 0x00180000) >> 19
        if mpeg_version == 0b01:
            # reserved value
            continue
        elif mpeg_version != 0b11:
            # not MPEG1 (= not MP3)
            continue

        # MPEG Layer: 0bxxxxxxxx xxxxxLLx xxxxxxxx xxxxxxxx
        mpeg_layer = (header_number & 0x00060000) >> 17
        if mpeg_layer == 0b00:
            # reserved value
            continue
        elif mpeg_layer != 0b01:
            # not Layer 3 (= not MP3)
            continue

        # CRC? 0bxxxxxxxx xxxxxxxC xxxxxxxx xxxxxxxx
        has_crc = (((header_number & 0x00010000) >> 16) == 0b1)

        # bitrate: 0bxxxxxxxx xxxxxxxx BBBBxxxx xxxxxxxx
        bit_rate_index = (header_number & 0x0000F000) >> 12
        if bit_rate_index == 0b0000:
            # weird bitrate
            continue
        elif bit_rate_index == 0b1111:
            # reserved value
            continue
        bit_rate = mp3_bit_rates[bit_rate_index]

        # sampling rate: 0bxxxxxxxx xxxxxxxx xxxxRRxx xxxxxxxx
        sampling_rate_index = (header_number & 0x00000C00) >> 10
        if sampling_rate_index == 0b11:
            # reserved value
            continue
        sampling_rate = mp3_sampling_rates[sampling_rate_index]

        # padding? 0bxxxxxxxx xxxxxxxx xxxxxxPx xxxxxxxx
        has_padding = (((header_number & 0x00000200) >> 9) == 0b1)

        # private bit is not interesting

        # channel mode: 0bxxxxxxxx xxxxxxxx xxxxxxxx CCxxxxxx
        channel_mode = (header_number & 0x000000C0) >> 6

        # mode extension: 0bxxxxxxxx xxxxxxxx xxxxxxxx xxEExxxx
        mode_extension = (header_number & 0x00000030) >> 4

        # copyright bit and original bit are not interesting

        # emphasis: 0bxxxxxxxx xxxxxxxx xxxxxxxx xxxxxxEE
        emphasis = (header_number & 0x00000003)
        if emphasis == 0b10:
            # reserved value
            continue

        # at this point, it's an MP3 file
        is_mp3 = True

        # calculate the frame length
        frame_length = floor(144 * bit_rate / sampling_rate)
        if has_padding:
            frame_length += 1

        # ker-BLAM
        last_mp3_bytes += my_bytes
        last_mp3_bytes += stream.read(frame_length - 4)
        
        # prepare for the next scan-read
        my_bytes = b"\0" + stream.read(3)

if __name__ == '__main__':
    import sys
    for arg in sys.argv[1:]:
        with open(arg, "rb") as f:
            find_mp3(f, arg)

@RavuAlHemio: for cases when you know there is only one mp3 file inside the binary it might also be possible to use ffmpeg -i mybinaryfile.bin -acodec copy fixed_file.mp3