Live transcribe

Would be great to see the trancrip being output while speaking instead of waiting until after. Maybe this could be done in chunks of e.g. 10 seconds but then again I know whisper uses context to improve accuracy so I am not sure I know the model too well to find my way around this problem.