feature request: multiple speaker detection

[EroticPlum]

the ability for the model to detect multiple voices and identify them with colour or something

additionally ability to export transcript with those voices separated.

I realise this is likely a big task however and may not be feasible