diff --git a/README.md b/README.md
new file mode 100644
index 0000000..0ee8d87
--- /dev/null
+++ b/README.md
@@ -0,0 +1,17 @@
+# Netease Cloud Music Copyright Protection File Dump (Python version)
+
+## 简介
+
+早就在算计网易云ncm文件，不过一直没什么机会，前些天突然看到已经有大佬 [anonymous5l/ncmdump ](https://github.com/anonymous5l/ncmdump)实现了，因为好奇加密算法就用Python移植了下，啊大佬不愧是大佬，感激！测试发现转换出来的媒体文件都已包含媒体信息，故未再实现原repo中的写tag操作。还有Python实现比C++慢不少。仅供学习交流，请勿传播扩散。用Python处理字节真让人头大emmm
+
+## 依赖
+
+```
+pip(3) install pycrypto
+```
+
+## 使用
+
+```
+python(3) ncmdump.py [files ...]
+```
\ No newline at end of file
diff --git a/ncmdump.py b/ncmdump.py
new file mode 100644
index 0000000..422da51
--- /dev/null
+++ b/ncmdump.py
@@ -0,0 +1,117 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Sun Jul 15 01:05:58 2018
+
+@author: Nzix
+"""
+
+import binascii
+import struct
+import base64
+import json
+import os
+from Crypto.Cipher import AES
+
+def dump(file_path):
+
+    core_key = binascii.a2b_hex("687A4852416D736F356B496E62617857")
+    meta_key = binascii.a2b_hex("2331346C6A6B5F215C5D2630553C2728")
+    unpad = lambda s : s[0:-(s[-1] if type(s[-1]) == int else ord(s[-1]))]
+
+    f = open(file_path,'rb')
+
+    # magic header
+    header = f.read(8)
+    assert binascii.b2a_hex(header) == b'4354454e4644414d'
+
+    # key data
+    f.seek(2, 1)
+    key_length = f.read(4)
+    key_length = struct.unpack('<I', bytes(key_length))[0]
+
+    key_data = f.read(key_length)
+    key_data_array = bytearray(key_data)
+    i = 0
+    while i < len(key_data_array):
+        key_data_array[i] ^= 0x64
+        i += 1
+    key_data = bytes(key_data_array)
+
+    cryptor = AES.new(core_key, AES.MODE_ECB)
+    key_data = unpad(cryptor.decrypt(key_data))[17:]
+    key_length = len(key_data)
+
+    # key box
+    key_data = bytearray(key_data)
+    key_box = bytearray(range(256))
+    c = 0
+    last_byte = 0
+    key_offset = 0
+
+    for i in range(256):
+        swap = key_box[i]
+        c = (swap + last_byte + key_data[key_offset]) & 0xff
+        key_offset += 1
+        if key_offset >= key_length:
+            key_offset = 0
+        key_box[i] = key_box[c]
+        key_box[c] = swap
+        last_byte = c
+
+    # meta data
+    meta_length = f.read(4)
+    meta_length = struct.unpack('<I', bytes(meta_length))[0]
+
+    meta_data = f.read(meta_length)
+    meta_data_array = bytearray(meta_data)
+    i = 0
+    while i < len(meta_data_array):
+        meta_data_array[i] ^= 0x63
+        i += 1
+    meta_data = bytes(meta_data_array)
+    meta_data = base64.b64decode(meta_data[22:])
+
+    cryptor = AES.new(meta_key, AES.MODE_ECB)
+    meta_data = unpad(cryptor.decrypt(meta_data)).decode('utf-8')[6:]
+    meta_data = json.loads(meta_data)
+
+    # crc32
+    crc32 = f.read(4)
+    crc32 = struct.unpack('<I', bytes(crc32))[0]
+
+    # album cover
+    f.seek(5, 1)
+    image_size = f.read(4)
+    image_size = struct.unpack('<I', bytes(image_size))[0]
+    image_data = f.read(image_size)
+
+    # media data
+    file_name = meta_data['musicName'] + '.' + meta_data['format']
+    m = open(os.path.join(os.path.split(file_path)[0],file_name),'wb')
+
+    chunk = bytearray()
+    while True:
+        chunk = bytearray(f.read(0x8000))
+        chunk_length = len(chunk)
+        if not chunk:
+            break
+
+        for i in range(chunk_length):
+            j = (i + 1) & 0xff;
+            chunk[i] ^= key_box[(key_box[j] + key_box[(key_box[j] + j) & 0xff]) & 0xff]
+
+        m.write(chunk)
+
+    m.close()
+    f.close()
+
+if __name__ == '__main__':
+    import sys
+    if len(sys.argv) > 1:
+        for file_path in sys.argv[1:]:
+            try:
+                dump(file_path)
+            except:
+                pass
+    else:
+        print('please input file path!')