WeChat audio converter script. See http://kronopath.net/blog/extracting-audio-messages-from-wechat/ for more details.

converter.py

# WeChat aud file converter to wav files
# Dependencies:
#   SILK audio codec decoder (available at https://github.com/gaozehua/SILKCodec)
#   ffmpeg
#
# By Gabriel B. Nunes ([email protected])
# Adapted from another script by Nicodemo Gawronski ([email protected])
# 

import os, argparse, subprocess
from datetime import datetime


def is_silk_file(file_path):
    # If the file has a SILK file header, we know it's a SILK file. Otherwise, we assume it's the
    # older AMR file format.
    input_file = open(file_path, 'rb');
    header = input_file.read(10)
    silk_header = '#!SILK_V3'
    # WeChat adds an extraneous byte at the beginning of its SILK files for some reason.
    input_file.close()
    return header[1:] == silk_header


def convert_amr_file(input_dir, input_file_name, output_dir):
    # These files are AMR files without the AMR header, so they can be converted by just adding the
    # AMR file header and then converting from AMR to WAV.

    input_file_path = os.path.join(input_dir, input_file_name);
    input_file = open(input_file_path, 'rb')

    intermediate_file_name = input_file_name.replace(".aud", ".amr")
    intermediate_file_path = os.path.join(output_dir, intermediate_file_name)
    intermediate_file = open(intermediate_file_path, 'wb')

    amr_header = "#!AMR\n"
    intermediate_file.write(amr_header + input_file.read())

    input_file.close()
    intermediate_file.close()

    output_file_name = input_file_name.replace(".aud", ".wav")
    output_file_path = os.path.join(output_dir, output_file_name)

    black_hole_file = open("black_hole", "w")
    subprocess.call(["ffmpeg", "-i", intermediate_file_path, output_file_path],
                    stdout = black_hole_file, stderr = black_hole_file)
    black_hole_file.close()
    
    # Delete the junk files
    os.remove("black_hole")
    os.remove(intermediate_file_path)


def convert_silk_file(input_dir, input_file_name, decoder_file_path, output_dir):
    # These files are encoded with the SILK codec, originally developed by Skype. They can be
    # converted by stripping out the first byte and then using the SILK decoder.

    input_file_path = os.path.join(input_dir, input_file_name);
    input_file = open(input_file_path, 'rb')

    intermediate_file_1_name = input_file_name.replace(".aud", ".silk")
    intermediate_file_1_path = os.path.join(output_dir, intermediate_file_1_name)
    intermediate_file_1 = open(intermediate_file_1_path, 'wb')

    # WeChat adds a single extra byte before their SILK files. We need to strip it out.
    intermediate_file_1.write(input_file.read()[1:])
    
    input_file.close()
    intermediate_file_1.close()

    intermediate_file_2_name = input_file_name.replace(".aud", ".pcm")
    intermediate_file_2_path = os.path.join(output_dir, intermediate_file_2_name)
    intermediate_file_2 = open(intermediate_file_2_path, 'wb')

    output_file_name = input_file_name.replace(".aud", ".wav")
    output_file_path = os.path.join(output_dir, output_file_name)
    
    black_hole_file = open("black_hole", "w")

    # Use the SILK decoder to convert it to PCM
    subprocess.call([decoder_file_path, intermediate_file_1_path, intermediate_file_2_path],
                    stdout = black_hole_file, stderr = black_hole_file)

    # And then ffmpeg to convert that to wav
    subprocess.call(["ffmpeg", "-f", "s16le", "-ar", "24000",
                    "-i", intermediate_file_2_path, output_file_path],
                    stdout = black_hole_file, stderr = black_hole_file)

    black_hole_file.close()
    intermediate_file_2.close()

    # Delete the junk files
    os.remove("black_hole")
    os.remove(intermediate_file_1_path)
    os.remove(intermediate_file_2_path)


class Main(argparse.Action):
    def __call__(self, parser, namespace, values, option_string=None):
        
        audio_src = values
        now = datetime.utcnow()
        now = datetime.strptime(str(now), '%Y-%m-%d %H:%M:%S.%f')
        now = now.strftime('%d-%m-%Y %H.%M.%S')
        converted = now + "_converted"
        os.mkdir(converted)

        try:
            # Decide which one of these are AMR files and which are SILK, and then convert.
            for dirname, dirnames, filenames in os.walk(audio_src):
                for filename in filenames:
                    if filename[0] == '.': continue
                    input_path = os.path.join(dirname, filename)
                    if (is_silk_file(input_path)):
                        convert_silk_file(dirname, filename, namespace.silk_decoder, converted)
                    else:
                        convert_amr_file(dirname, filename, converted)

            print "Done!"
        except:
            print("Something went wrong converting the audio files.\n"
                "Common problems:\n"
                "You may be missing the dependencies (ffmpeg and/or the SILK codec decoder).\n"
                "The decoder (and its dependencies) must be in the specified path.\n"
                "The SILK codec decoder also can't handle very large file paths.\n"
                "Try a shorter path to your input directory.")


parser = argparse.ArgumentParser(description=".aud converter: convert wechat .aud files into .wav",
        epilog="This script is an open source tool under the GNU GPLv3 license. Uses content "\
                "modified from a tool originally for DEFT 8.")
parser.add_argument("Folder", action=Main, help=".aud files root folder.")
parser.add_argument("-s", "--silk-decoder", nargs="?", default="./decoder",
        help="Path to the SILK codec decoder program.")

args = parser.parse_args()

Hello!
I found this script just now and used it to extract audio from the latest version of wechat (Android, wechat v.6.6.7), but had to make some changes to get it to work. I realize this project may never be updated, but want to add this comment as a reference for others who may need it.

1. The latest wechat changed the extension from .aud to .amr, format remains the same, so we need to change the input extension:
   Line 62: Change to intermediate_file_1_name = input_file_name.replace(".amr", ".silk")
   Line 72: Change to intermediate_file_2_name = input_file_name.replace(".amr", ".pcm")
   Line 76: Change to output_file_name = input_file_name.replace(".amr", ".wav")

2. ffmpeg prompts to overwrite, which is suppressed by the stdout/stderr overrides, so the script appears to hang but is just waiting for input. Add forced overwrite:
   Line 86: Change to subprocess.call(["ffmpeg", "-y", "-f", "s16le", "-ar", "24000",

3. The latest python (3.7.0) has a different handling of byte/string compare/concat, so we need to encode strings to bytes before using them:
   Line 19: Change to silk_header = str.encode('#!SILK_V3')
   Line 36: Change to amr_header = str.encode("#!AMR\n")

4. Small typo at the very end:
   Line 120: Change to print ("Done!")

Hope this helps somebody out there who's having trouble, thanks for creating this script and documenting the process!

Thanks @Teslapirate, that helped me get a SILK file out but couldn't get it converted to wav, then I found this.

Then after cloning run chmod +x converter.sh and then ./converter.sh <silkfile> wav converted them to WAV files.

(Linux)

Hello! I found this script just now and used it to extract audio from the latest version of wechat (Android, wechat v.6.6.7), but had to make some changes to get it to work. I realize this project may never be updated, but want to add this comment as a reference for others who may need it.

1. The latest wechat changed the extension from .aud to .amr, format remains the same, so we need to change the input extension:
   Line 62: Change to intermediate_file_1_name = input_file_name.replace(".amr", ".silk")
   Line 72: Change to intermediate_file_2_name = input_file_name.replace(".amr", ".pcm")
   Line 76: Change to output_file_name = input_file_name.replace(".amr", ".wav")
2. ffmpeg prompts to overwrite, which is suppressed by the stdout/stderr overrides, so the script appears to hang but is just waiting for input. Add forced overwrite:
   Line 86: Change to subprocess.call(["ffmpeg", "-y", "-f", "s16le", "-ar", "24000",
3. The latest python (3.7.0) has a different handling of byte/string compare/concat, so we need to encode strings to bytes before using them:
   Line 19: Change to silk_header = str.encode('#!SILK_V3')
   Line 36: Change to amr_header = str.encode("#!AMR\n")
4. Small typo at the very end:
   Line 120: Change to print ("Done!")

Hope this helps somebody out there who's having trouble, thanks for creating this script and documenting the process!

I just used this now and didn't have to change anything except adding -y to the params. iOS backup from today, everything else downloaded and compiled today. I also made it output to a directory per conversation. instead of one huge folder of the same filenames.

I'll see if I can do a pull req on this.