is it possible to map people to transcript with AWS transcribe?

Question

I need to use some transcription service to transcribe speech. I recieve very good results with AWS transcribe. But i also need to know which person says what. Is this possible with AWS transcribe? I cannot find anything about this.

ex. what i get now is

Speaker 1: hello how are you?

Speaker 2: I am good how are you?

Speaker 1: fine...

What I would like to get is to be able to know who is speaker 1 and two. for example by uploading samples

Bob: hello how are you?

Alice: I am good how are you?

Bob: fine...

is this possible with AWS?

Answer 1

There are two approaches:

• Use multiple channels (eg left & right audio) to separate speakers. This is good for telephone calls with two distinct speakers. See: Transcribing multi-channel audio - Amazon Transcribe
• Identifying speakers with diarization, see: Identifying speakers (speaker diarization) - Amazon Transcribe

These will identify separate speakers. However, you cannot upload a "sample" and then identify the specific speaker.