Over lapping audio/Cross talk #512
Description
When two people speak together (like in an interview), the model does very well to capture the words spoken individually by both speakers but the timestamps assigned to those words are aligned so that there is no overlap of timestamps. Model decides to assign one person a set of start and end times which are bunched together incorrectly and the second person's word's timestamps start after that -- instead of what happened which is both speakers having the same time stamps when they speak over each other.
Is it possible to force that? That we get overlapping timestamps instead of in sequence time stamps?