Recordings in Stereo

[it26]

Does anyone know if 3CX supports recordings in stereo, or a solution for doing so? We are working on an AI solution that can transcribe calls a seperate callers by channel. Thank you.

 

[Charles_3CX]

The audio format of call recordings is as shown below.

Currently, 3CX does not support stereo recordings. However, you can use one of the supported AI transcription services built into the system: 3CX Transcription (included free with an Enterprise license), Google Speech-to-Text via the Google Bucket integration, or OpenAI.

 

[it26]

Do you know if the transcription services are able to separate the transcriptions between Caller1 and Caller2?

 

[SteveITS]

I think this is coming next (?) but yes: https://www.3cx.com/phone-system/ai-transcription/

 

[it26]

Do you know where documentation about setting this up would be? Thank you very much. This is exactly what I was looking for.

 

[VoIPTools]

While AI is getting better every day, diarization from a mono recording (today) has some limitations. Some models are specifically designed and tuned to do diarization. OpenAI is typically rated at about 80-90% accuracy with some caveats. The following can impact that accuracy estimate:

• The number of speakers beyond two will impact the accuracy
• The similarity of speaker voices
• Frequency of when two or more speakers talk at the same time
• Noise in the background -- laughing, music, quality of the recording, etc.

Nevertheless, compared to just a couple of years ago, AI has made amazing advancements in transcription. Our company has been transcribing 3CX recordings for 15+ years and I can honestly say this new technology is amazing. It's not perfect (yet) but it is getting closer to perfection on what seems like a daily basis.

 

[KyriacosS_3CX]

Do you know where documentation about setting this up would be? Thank you very much. This is exactly what I was looking for.

This will be a feature of the next update (U8) and documentation as well as the actual options in the PBX will be available once the update is out. An Alpha version for testing should not be too long from now.

 

[Steven Vu]

Just because I was looking for this for a while and think it might help someone else. It's not true multichannel and doesn't describe what happens when there's more than 2 speakers. Also doesn't say whether the agent is always on the L or R channel and whether it flips depending on inbound or outbound. If anyone wants to do some research, be my guest.

In Advanced > Parameters:

MS_CALLRECORDING_STEREO

Enable stereo call recording

 

[Alphabetic]

Just because I was looking for this for a while and think it might help someone else. It's not true multichannel and doesn't describe what happens when there's more than 2 speakers. Also doesn't say whether the agent is always on the L or R channel and whether it flips depending on inbound or outbound. If anyone wants to do some research, be my guest.

In Advanced > Parameters:

MS_CALLRECORDING_STEREO

Enable stereo call recording

All this does is enable this option.
you still require recommended licensing to utilise 3CX transcription services.

 

[Tonnie P]

Just because I was looking for this for a while and think it might help someone else. It's not true multichannel and doesn't describe what happens when there's more than 2 speakers. Also doesn't say whether the agent is always on the L or R channel and whether it flips depending on inbound or outbound. If anyone wants to do some research, be my guest.

In Advanced > Parameters:

MS_CALLRECORDING_STEREO

Enable stereo call recording

Is this parameter removed now, or do you need Enterprise/AI functions for this parameter to exist?
We are currently looking for ways to automate transcibe and summarize calls on our servicedesk and attatch them to the ticket, it looks like the integrated transcribe functions from 3cx is a bit to limted, so i was thinking to save these recordings (in stereo) and use something like n8n to create the workflow of transcribing, summarizing and sending it to our systems, having stereo recordings would improve things a lot.

 

[bitn2]

Is this parameter removed now, or do you need Enterprise/AI functions for this parameter to exist?
We are currently looking for ways to automate transcibe and summarize calls on our servicedesk and attatch them to the ticket, it looks like the integrated transcribe functions from 3cx is a bit to limted, so i was thinking to save these recordings (in stereo) and use something like n8n to create the workflow of transcribing, summarizing and sending it to our systems, having stereo recordings would improve things a lot.

Correct, this is an enterprise feature.