A way to merge speakers

When speaker detection acts suboptimal, sometimes there are many speakers detected, which are actually the same speaker.

At the moment one need to change every detected speaker change manually. If one could merge speakers this would be **way** faster to fix.