We have had quite a few customers asking for automated call transcription reports to be email directly to a nominated email address after every call.
We have lost a customer recently to a system that automatically sends.
Let us know if this is something that is on the roadmap.
While this is not built in at the moment, you can use a post call recording script to process and send an email.
I am not sure how to do this, do you have instructions you can share?
Depending on the hardware, add a RTX3000 series (RTX3090 is good/best) GPU, Whisper large v3, n8n to tie it all together, and you can have a transcript of every call emailed or (as we have done) uploaded to your CRM. We took it a step further and used our local AI LLM to create a summary of the call and upload that to our CRM too. The summary is still a work in progress to get all the fine details desired, albeit right out of the box it’s (the local LLM) been relatively good using sub $5000 worth of hardware (plus it does a lot more than summaries, it processes documents etc…)
You don’t this this whole setup, you just need a minimal script to ship the call recording to an AI service, retrieve the transcription and email it.
My small office is working on this functionality currently. We added transcripts to voicemail emails yesterday (calls are already transcribed), and it only took a few hours of coding/testing. It remains to be seen if further editing/coding is required, albeit so far so good.
Using a python script, open source transcribing engine, a RTX 3070 GPU, and one edit to the already working internal voicemail email GUI settings and it was working (admittedly beta for maybe another week or two to confirm no issues). The best part is for our small system, it added less than two seconds for the voicemail delivery which included the actual recording, so it’s a big win considering the cost of hardware vs increase in productivity if not having to listen to voicemails. (Ie, a 15-20 second VM takes about two seconds to transcribe)
One other note that comes to mind, for us, the computer requirements for transcribing is nothing compared to summaries, and having an outside API transcribe service is maybe/likely more difficult and less reliable to implement and maintain than internally.
PLUS, it may be HIGHLY advisable to receive consent from clients before exposing phone calls and their transcripts to a third-party, a well as being listed as an additional insured on the API transcribers insurance policy.
Today’s add on is taking the transcripts and summaries and updating the CRMs (three) we use. Well, by today, I mean it starts today and if it goes well, it’s working but the end of the day(likely will, it’s not difficult in theory).
Then back to summaries and tweaking the prompts and or LLM model to produce desired summaries(or using if necessary outside API, which likely adds a few cents to each call that we would like to avoid if we can)
Thank you, that makes sense, do you have a specific script to achieve this? Sorry, this is a bit technical for me.
With a lot of help from OpenAI, I was able to create a script that does the transcription and send the email, but to get it to work without sending dual emails for every voicemail I had to empty the email field for each user and create a csv file the script references for extension and email addresses. I also had to create a different service to watch for new wav files in (only) the Inboxes for each extension. Works sop far, but I wish this was something our friends at VitalPBX would make available to us via the GUI on small servers that aren’t big enough for the licensing they currently require.
In the past 2 weeks I have had 5 different companies ask for call recording to ai transcription and then to get results by email or web page. So this is now becoming a common request!!