Feature Request: Voicemail Transcription using Mozilla Single Binary called Whisperfile

[johnny]

I just found this article on Phoronix regarding Mozilla's new Whisperfile and decided to test it to transcribe some voicemails and it did exceptionally well locally using very little CPU usage a`t a very fast rate. Its a single binary that can Transcribe Audio Files like voicemails at very good accuracy.

I know FusionPBX has voicemail transcription where it would send the audio off to OpenAI to process and send transcribed voicemail in an email. I think this would be a better more private approach and its a simple 1 file binary which would make install and updating super easy.

I would love to have this implemented in DjangoPBX ;-)

Here's the link I follow

Mozilla Developing Whisperfile For Local Audio-To-Text Translation - Phoronix

www.phoronix.com

Here's the Git Page

Mozilla/whisperfile · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co

 

[Adrian Fretwell]

Thank you @johnny I will take a look at that. On the face of it, it does sound promising.

Te Voicemail system on DjangoPBX is going through a pre-beta overhaul. It originally used Freeswitch mod_voicemail, but we have seen a good few feature requests that have required that we build our own.

 

[johnny]

Thats sounds fantastic @Adrian Fretwell cant wait to test it =]

 

[pbxgeek]

I'm definitely adding this to FS PBX.

 

[pbxgeek]

@johnny Did you observe how much CPU it takes to run this locally by any chance? I'm curious because we can't afford to use all of the valuable CPU speed on AI and have the calls suffer from it. Does it take a lot to run?