Speech to text translation, how to implement in asterisk?

Question

Asterisk

There is CRM Vtiger integrated with asterisk, records of telephone conversations are attached to contacts, but I want to translate the conversation into text, there is a Google and Yandex API for this, but how to separate the phrases of the client and the manager?
Has anyone implemented something like this? share your experience.

Answer 1

Most likely, you use the MixMonitor command for recording on Asterisk, which writes both Call-legs to one file at once. Accordingly, for recognition it is necessary to write to two different files. This can be done with the Monitor command.
True, it is not clear how to attach them later to CRM ...
In general, you need to delve into your Asterisk and look for a compromise.

Answer 2

A split record made via Monitor, then you need to merge the channels via sox, for example.
https://www.nesono.com/node/275

Answer 3

It turns out that you need to write 2 different files, then send them to Google, since Yandex does not yet work with large files, and in general it still has a lot to do, and then get an answer and apply, it turns out 3 rubles a minute, not so sour, in Bitrix is ​​also implemented, apparently, and also 3 rubles a minute.