I think the key distinction is that there is no specific training data for that speaker. You can view the input as just the input voice to clone, not training e

I think the key distinction is that there is no specific training data for that speaker. You can view the input as just the input voice to clone, not training examples. It would be more like training examples if you had to give it specific phrases.