Transcribe is identifying speakers 0, speaker 1 as same person.

I assume that you are calling the StartTranscriptionJob API.

Check bellowing parameters:

• MediaFormat, MediaSampleRateHertz setting, if you have a specific one.
• MaxSpeakerLabels, Specify the maximum number of speakers you want to identify in your media. Note that if your media contains more speakers than the specified number, multiple speakers will be identified as a single speaker. If you specify the MaxSpeakerLabels field, you must set the ShowSpeakerLabels field to true.

Reference: StartTranscriptionJob

If it's a streaming one, check Identifying speakers in real-time streams

I'm facing a similar issue with the StartMedicalScribeJob. In the audio files I've tested, I've had two speakers in them but it's showing the same label for both the speakers in the transcript generated. Anything else need to be explicitly set for the service to properly distinguish between the speakers? My cuurent parameters are set like this: ShowSpeakerLabels = True MaxSpeakerLabels= 3 ChannelIdentification= False