Comment by ericmcer

can a model like this be used to replicate a loved one’s voice? isn’t it a model trained on audio instead of text? So synthesizing another voice would involve rewriting the weights?