Whisper AI - Option to change endpoint to self hosted

[Mark_ACA]

Really pleased to see whisper AI integration for transcribing call recordings.
However Whisper AI is pretty lightweight, free and easy to self host rather than paying OpenAI for the API

Would it be possible on a future release to allow integration of a self hosted instance which can be easily setup with the following ?.....

https://github.com/innovatorved/whisper.api

TIA

 

[VoIPTools]

3CX announced that they will be doing precisely this -- hosting their own "paid" transcription service. It's free until the end of this year, but it sounds like it will be a paid add-on service after the start of the year.

https://www.3cx.com/blog/releases/v20u4-call-transcription/

We can do it for you. DM me. Take a look at 3CX Recording Manager and 3CX Voicemail Manager. While our tools are not free, for very large implementations it would be far more cost effective than paying OpenAI. We will have to wait and see what 3CX's service costs.

I'm having staff try out your link right now.

 

[Mark_ACA]

Ah, that's not quite what i meant by self host. But i guess this will disincentivise them to allow using our own self hosted endpoint in future as they will get some revenue from theirs.

We already have our own AI server setup and use it for voicemails after they import to our CRM (as well as other functions in our CRM), I was hoping I could plug the recordings translation into this directly. I would have to do it the other way around, have the AI server pull recordings from the 3cx server periodically via API, and save the translations to a local database.

At least it's 'affordable' so lets see what the rates are Thanks.

 

[VoIPTools]

Our tools monitor for new recordings / voicemails in real-time and currently use OpenAI to perform the transcription. It would be easy to change our tools to use your suggested self-hosted whisper. Given that you are already doing it for voicemails, it wouldn't be hard to expand to recordings.

But this whisper implementation does not include analysis for things like agent scoring, identifying action items, summaries, and all the things we currently do with OpenAI, That said, you could use this project for transcription and OpenAI for everything else. It's the transcription step that is relatively expensive from OpenAI. If you have suggestions for a self-hosted option for the analysis, I can look at that too.

Look at our tools. We can give you some good ideas of what is possible, and I'm always looking for new uses of AI.