Transcription not sent in emails first time, is sent on resend.

[Davesworld]

I got transcription working with OpenAI, but the first emails sent out do not have transcription in them, only the attached files. When I go to the messages via the web GUI, they are transcribed, and if I resend the email manually, the transcription IS in the email.

 

[whut]

I have seen openai transcription in emails not be included far too often. The transcription does not seem to be returned quickly enough to be included in the email.

 

[Davesworld]

I have seen openai transcription in emails not be included far too often. The transcription does not seem to be returned quickly enough to be included in the email.

I figured it was timing. If only there were a way to delay the email composition.

 

[wouam31]

Hi,
Maybe extending the email sending time could fix the delay for the OpenAI information to arrive.
In the Email Queue, increase the interval to 300.

Have a great day.

 

[Davesworld]

Hi,
Maybe extending the email sending time could fix the delay for the OpenAI information to arrive.
In the Email Queue, increase the interval to 300.

Have a great day.

I don't see how that would work as the queue stores and delays what has already been composed as far as I know, in this case, prematurely but I just set it at 300 to see if it has any effect.

 

[Davesworld]

Hi,
Maybe extending the email sending time could fix the delay for the OpenAI information to arrive.
In the Email Queue, increase the interval to 300.

Have a great day.

It had no effect as I suspected since it sends the mail to the queue immediately and the delay is just the send interval of what has already been composed and queued. It was worth a shot though.

I am curious as to why there are two sections, speech and transcribe. In the speech resources, there exists both eleven labs and openai. In the transcribe resources, I see azure, google, openai and watson. Openai is the only one that is in both places.

 

[Davesworld]

I have seen openai transcription in emails not be included far too often. The transcription does not seem to be returned quickly enough to be included in the email.

I've only had a few that were transcribed in time for the mailer. It sends it to the queue a bit too fast.

 

[whut]

It had no effect as I suspected since it sends the mail to the queue immediately and the delay is just the send interval of what has already been composed and queued. It was worth a shot though.

I am curious as to why there are two sections, speech and transcribe. In the speech resources, there exists both eleven labs and openai. In the transcribe resources, I see azure, google, openai and watson. Openai is the only one that is in both places.

Transcribe = speech to text
Speech= text to speech

Some services only provide 1 of these functions. Some functions use different settings (ex: API keys) if they provide both functions. There are more reasons why the services/applications are individualized.

It seems that the email needs to wait for the transcription payloads to be returned before sending.

 

[Davesworld]

Transcribe = speech to text
Speech= text to speech

Some services only provide 1 of these functions. Some functions use different settings (ex: API keys) if they provide both functions. There are more reasons why the services/applications are individualized.

It seems that the email needs to wait for the transcription payloads to be returned before sending.

Thanks. Hopefully a parameter can be added to delay the email composition once a wave file is recorded or OpenAI needs to be faster.

 

[Davesworld]

Transcribe = speech to text
Speech= text to speech

Some services only provide 1 of these functions. Some functions use different settings (ex: API keys) if they provide both functions. There are more reasons why the services/applications are individualized.

It seems that the email needs to wait for the transcription payloads to be returned before sending.

It makes sense now since the only one that is shown in both categories is OpenAI.

 

[whut]

It makes sense now since the only one that is shown in both categories is OpenAI.

openai is the default choice for both the text to speech and the speech to text modules. If you look at https://github.com/fusionpbx/fusionpbx-app-transcribe/tree/main/resources/classes

and https://github.com/fusionpbx/fusionpbx-app-speech/tree/main/resources/classes

you will see the currently available providers.