Record Mic in and RTP packet streams in a stereo wav file

[mjalil]

Hi dear @sipsorcery Aaron,
I'm trying to modify the Softphone example to record both sending and receiving voices of a SIP call.
I used the RecordCall Example to record calls, but unfortunately, the recorded wave file only contains the voice of the calling side, not my voice (microphone)

I also Googled it and used AI chat to find out how to do it. Then, I searched for this in the issues, documentation, and project articles with no success.
They proposed to directly use NAudio WaveInEvent to get raw samples of the microphone and write them to a wav file, and use RtpPacket to write another wav file, then merge them using NAudio in a stereo wav file.
May you please guide me, if there is a better solution?

[sipsorcery]

There's not a great out of the box solution. There are two separate audio streams involved, the one you're sending and the one you're receiving. To combine them properly into a single recording you need an audio mixer. NAudio does have a mixer but there are no examples in the sipsorcery code base.

The mixer aside the missing link in the softphone example is getting the samples for the audio stream being sent, which typically corresponds to the microphone captures. The encoded samples can be hooked via the OnAudioSourceEncodeSample. If the encoding being used is PCM then the samples can be decoded in the same manner as the RecordCall example and written to a .wav file, or if you get a mixer wired up written directly to it so you have a single .wav file for both streams.

In theory the WindowsAudioEndPoint could generate raw samples via the OnAudioSourceRawSample but it's not currently wired up (it wouldn't be that big a job to do so).

[ElDuderinoBerlin]

There's not a great out of the box solution. There are two separate audio streams involved, the one you're sending and the one you're receiving. To combine them properly into a single recording you need an audio mixer. NAudio does have a mixer but there are no examples in the sipsorcery code base.

It's indeed a quite a decent solution. What you hear is what you get. I did this with CSCore not NAudio with Sipsorcery. I wrote my own SIPSorceryMedia.Windows.CSCore library for that.
Going for the audio APIs of the OS frees you from handling RTPs etc. and you get no cracks and you can record and mix with at any rate and audio format.

This works not without audio hardware like sound card or headset with builtin audio driver.