[janniks]

Would it be possible to mix multiple people/voices in a training set? Or does that confuse the AI? Could be interesting to create a real but kind of not real model. If that makes sense…