I wonder if it would be possible to use a voice sample from the therapist to ensure that the AI recognizes the therapist's voice rather than the client's. It's too frequent that my voice gets translated into comments/feedback that the Ai is thinking is said by the client.