Convert raw PCAP files into a .wav file

pcap2wav

#!/bin/bash
#
# pcap2wav
# Original Author: Michael Collins <msc@freeswitch.org>
#Standard disclaimer: batteries not included, your mileage may vary...
# Updated by Avi Marcus <avi@bestfone.com>
#
# Accepts arg of pcap file w/only 2 RTP streams
# Creates a .<codec> file and a .wav file
# For codecs other than PCMA and PCMU the script calls fs_cli and does a little recording to create the wav file(s)
# Current codec support: g711a/u, GSM, G722, G729

# check for -h -help or --help

if [[ $1 == "-h" || $1 == "-help" || $1 == "--help" || $1 == "" ]]
then
    cat <<EOF

pcap2wav is a simple utility to make it easier to extract the audio from a pcap

Dependencies:
   apt-get install -y tshark sox
   yum install wireshark sox
   
Usage:

  pcap2wav [opts] filename.pcap [target filename]

Script attempts to create a few files: a .<codec> file and a .wav file for each RTP stream

It requires Tshark to be installed on the system. If a codec other than PCMA or PCMU
 is used then the script will attempt to use fs_cli to decode and create a wav.

Supported codecs:
 PCMU (G711 ulaw)
 PCMA (G711 Alaw)
 GSM
 G722 (requires fs_encode)
 G729 (requres fs_encode with mod_com_g729)

Supported options:
 -z  Perform "clean and zip" - After converting to wav files the program will "clean up"
                               by putting the wav files into a .tgz file and then removing
                               the .wav and .<codec> files from the disk.
EOF

exit
fi

if [[ $1 == "-z" ]]
then
    CLEAN=true
    CAPFILE=$2
    TARGETFILE=$3
else
    CLEAN=false
    CAPFILE=$1
    TARGETFILE=$2
fi

LOGDIR=/var/log
TSHARK=`which tshark`
SOX=`which sox`
FSENCODE=`which fs_encode`

if [[ $TSHARK == "" ]]
then
    echo "Tshark not found. Please install Tshark and then re-run this script"
    exit
fi

if [[ $SOX == "" ]]
then
    echo "Sox not found. Please install Sox and then re-run this script"
    exit
fi

# Make sure pcap exists
if [ -f $CAPFILE ]
then
    echo "Found $CAPFILE, working..."
else
    echo "$CAPFILE not found, exiting."
    exit
fi

# Set target file names; default is "pcap2wav.<codec>" and "pcap2wav.wav"
if [[ $TARGETFILE == "" ]]
then
    TARGETFILE="/tmp/pcap2wav"
else
    echo "Using $TARGETFILE"
fi

echo "Checking $CAPFILE for RTP streams..."

# Locate RTP streams, put into temp file
tshark -n -r $CAPFILE -Y rtp -T fields -e rtp.ssrc -e udp.dstport -Eseparator=, | sort -u > /tmp/pcap2wav.tmp

# Count the RTP streams
num_streams=`grep -c "" /tmp/pcap2wav.tmp`
streams=( $(cat /tmp/pcap2wav.tmp) )

#Extract stream ssrc and port
for item in `seq 1 $num_streams`; do
    index=$((item-1))
    ssrc[$item]=`echo ${streams[$index]} | cut -d, -f1`
    port[$item]=`echo ${streams[$index]} | cut -d, -f2`
done

payload_type=`tshark -n -r $CAPFILE -T fields -e rtp.p_type | grep -P '\d+' | head -n 1`
case $payload_type in
    0) codec='PCMU'
        for item in `seq 1 $num_streams`; do
            convert[$item]="$SOX -t ul -r 8000 -c 1 ${TARGETFILE}_$item.$codec ${TARGETFILE}_$item.wav"
        done
        ;;
    3) codec='GSM'
        for item in `seq 1 $num_streams`; do
            convert[$item]="$SOX -t gsm -r 8000 -c 1 ${TARGETFILE}_$item.$codec ${TARGETFILE}_$item.wav"
        done
        ;;
    8) codec='PCMA'
        for item in `seq 1 $num_streams`; do
            convert[$item]="$SOX -t al -r 8000 -c 1 ${TARGETFILE}_$item.$codec ${TARGETFILE}_$item.wav"
        done
        ;;
    9) codec='G722'
        for item in `seq 1 $num_streams`; do
            convert[$item]="$FSENCODE ${TARGETFILE}_$item.$codec ${TARGETFILE}_$item.wav"
        done
        ;;
    18) codec='G729'
        for item in `seq 1 $num_streams`; do
            convert[$item]="$FSENCODE -l mod_com_g729 ${TARGETFILE}_$item.$codec ${TARGETFILE}_$item.wav"
        done
        ;;
esac

if [ -z "$codec" ]; then
    echo "Unable to determine codec from payload type: $payload_type"
    exit
fi

echo "Target files to create:"
for item in `seq 1 $num_streams`; do
    echo "${TARGETFILE}_$item.$codec and ${TARGETFILE}_$item.wav"
done

echo
for item in `seq 1 $num_streams`; do
    echo "Stream $item ssrc / port: ${ssrc[$item]} / ${port[$item]}"
done
echo

for item in `seq 1 $num_streams`; do
    echo "Extracting payloads $item from ${ssrc[$item]}..."
    tshark -n -r $CAPFILE -Y "rtp.ssrc == ${ssrc[$item]}" -T fields -e rtp.payload > /tmp/pcap2wav.payloads${item} 2> /dev/null
    for payload in `cat /tmp/pcap2wav.payloads${item}`;do IFS=:;for byte in $payload; do printf "\\x$byte" >> ${TARGETFILE}_$item.$codec; done; done
    unset IFS
    command="${convert[$item]}"
    $command
done

# If two streams then assume they're a pair and combine them nicely
if [[ $num_streams == "2" ]]
then
    echo "Combining 2 streams into a single wav file for convenience"
    # Find shorter recording, calc time diff in samples
    samples1=`soxi -s ${TARGETFILE}_1.wav`
    samples2=`soxi -s ${TARGETFILE}_2.wav`

    if [[ $samples1 -gt $samples2 ]]
    then
        longer="${TARGETFILE}_1.wav"
        shorter="${TARGETFILE}_2.wav"
        delay=`expr $samples1 - $samples2`
    else
        longer="${TARGETFILE}_2.wav"
        shorter="${TARGETFILE}_1.wav"
        delay=`expr $samples2 - $samples1`
    fi

    pad="${delay}s"
    command="$SOX $shorter ${TARGETFILE}_tmp.wav pad $pad 0s"
    $command

    # Create "combined" file, padding beginning with silence
    command="$SOX -m ${TARGETFILE}_tmp.wav $longer ${TARGETFILE}_mixed.wav"
    $command
    rm -fr ${TARGETFILE}_tmp.wav

fi

if [[ $CLEAN == "true" ]]
then
    echo "Clean option"
    ZIPFILE=${TARGETFILE}.tgz
    rm -fr $ZIPFILE
    /bin/tar czf $ZIPFILE ${TARGETFILE}*wav > /dev/null 2>& 1
    for item in `seq 1 $num_streams`; do
        rm -fr ${TARGETFILE}_$item.*
    done
    rm -fr $TARGETFILE.tmp
else
    echo "No clean option specified - leaving .<codec> and .wav files on system."
fi

echo
echo "Operation complete"
echo

is there any way to export wav file starting after "200 connect"?

Can you explain this line?

payload_type=tshark -n -r $CAPFILE -T fields -e rtp.p_type | grep -P '\d+' | head -n 1

my versions of grep don't seem to have a -P flag.

@craigbruenderman try replacing -P with -E. That worked for me on macOS. You basically are telling grep to use extended regular expressions to search for a pattern. In this case the pattern being one or more digits.

Is this script massively broken for anyone else? Ubuntu has no package for fs_encode, but the script's error handling happily plows forward regardless, leading to the output files being painful, high-pitched ringing noises at full volume.

Is this script massively broken for anyone else? Ubuntu has no package for fs_encode, but the script's error handling happily plows forward regardless, leading to the output files being painful, high-pitched ringing noises at full volume.

Interesting, there's no check here that fs_encode exists.

It's only for G722 and G729 (requires licenses for g729).

It's installed via the freeswitch package, see the latest install instructions for debian 10: https://freeswitch.org/confluence/display/FREESWITCH/Debian+10+Buster

Hi avimar,
i tried your script and i got an error here -
/usr/bin/sox FAIL formats: can't open input file `outputofpcap_3.PCMA': No such file or directory

$ ./pcap2wav -z test.pcap outputofpcap.wav
Found test.pcap, working...
Using outputofpcap.wav
Checking test.pcap for RTP streams...
Target files to create:
outputofpcap.wav_1.PCMA and outputofpcap.wav_1.wav
outputofpcap.wav_2.PCMA and outputofpcap.wav_2.wav
outputofpcap.wav_3.PCMA and outputofpcap.wav_3.wav

Stream 1 ssrc / port: 0x37968baa / 40392
Stream 2 ssrc / port: 0x3796cb71 / 40392
Stream 3 ssrc / port: 0xaaaaaaaa / 40392

Extracting payloads 1 from 0x37968baa...
Extracting payloads 2 from 0x3796cb71...
Extracting payloads 3 from 0xaaaaaaaa...
/usr/bin/sox FAIL formats: can't open input file `outputofpcap.wav_3.PCMA': No such file or directory
Clean option

Operation complete

Pls, check.

All 3 tshark commands in the script need an additional option in order to work for me with raw RTP pcap file without conversation:
-o rtp.heuristic_rtp:TRUE

If this can help anyone... I had some problems with this part:

tshark -n -r $CAPFILE -Y "rtp.ssrc == ${ssrc[$item]}" -T fields -e rtp.payload > /tmp/pcap2wav.payloads${item} 2> /dev/null
for payload in `cat /tmp/pcap2wav.payloads${item}`;do IFS=:;for byte in $payload; do printf "\\x$byte" >> ${TARGETFILE}_$item.$codec; done; done
unset IFS

Because tshark returns the output of rtp.payload as hexa string not separated by colons (like "555555" instead of "55:55:55").
To fix this, I replaced the above snippet by this one using xxd:

tshark -n -r $CAPFILE -Y "rtp.ssrc == ${ssrc[$item]}" -T fields -e rtp.payload 2> /dev/null | xxd -r -p > ${TARGETFILE}_$item.$codec

Here is my version of tshark:

$ tshark --version
TShark (Wireshark) 3.2.3 (Git v3.2.3 packaged as 3.2.3-1)

Has anyone else noticed this issue?
One more thing, where can I retrieve fs_encode for g729?

Any idea how to change this script in order for it to take into consideration silence suppression? When any of the streams has silence suppression both streams are unsynchronized after that.